International Relations and Defence Committee
Wednesday 16th September 2026
Skip to latest contribution
This debate has concluded
(began 1 month, 3 weeks ago)
Watch LiveSkip to latest contribution
This debate has concluded
10:33
Lord Robertson of Port Ellen
-
-
Lord Robertson of Port Ellen
Good morning and welcome to the International Relations and Defence Committee of the House of Lords. Today we have been looking at the multilateral system and specifically now we are looking at how it relates to to artificial intelligence. I've received apologies from Lord Houghton and Lord Darroch. Our special adviser, Andrew Erhart, is in Boston today and is in the middle of the night. Clearly joining us. By by Zoom.
I thank you very much for making yourselves available today for our interrogation and , and addressing this subject of how the multilateral institutions might deal with the problems and difficulties of AI. So I remind you and others that this session is being streamed live on Parliament TV. Transcript will be taken and you will be able to correct any minor changes in in that as well. You'll get a copy of the transcript.
And I remind members that if they have relevant interests, they should declare them. And I declare that I am a senior counsellor with the Washington based Cohen Group. Right at the at the very beginning. So if I could if I could ask the first question. As you know, the the committee is looking at the whole multilateral system as it applies to the United Kingdom and hopefully will make recommendations in relation to to the UK's policy in relation to the various multilateral institutions that exist in the world today.
So that is why we are looking at the whole question of artificial intelligence in the context of the the global picture and the way in which multilateral institutions have already or may in the future, look at these things. So so what are the key governance issues related to AI that that multilateral engagement really needs to address? So if I if I could ask Mr. DOS if he if he would like to ask the answer the question first.
Certainly it's a real pleasure and privilege to be here. There's such a broad list of governance challenges. I'll pick out just a few of them. I think the first would be safety . Agreeing common methods for evaluating advanced AI, sharing risk information and responding to serious incidents. And we've seen in the news just the last few days the importance of Agentic AI and frontier capabilities in the cyber domain.
The second would be international security. From cyber attacks to AI enabled biological threats. Military and more. More generally. Third, regulatory interoperability. There's different approaches in the EU, China, US, UK and so on. Those differences are legitimate, but we need to make sure our systems are remain compatible. The next one, and this is often neglected I think, but is increasingly important, is cultural and linguistic diversity, that AI must not be just for greater efficiency, but it also needs to to reflect the wisdom of of humanity as a whole.
Information integrity. So coordinating responses to election interference that AI can supercharge science and public health. There's huge potential as reasoning models develop and through eventually recursive self-improvement, to have a cascade of new discoveries in mathematics, science, and medicine. And I think governance is important to harness that in the most effective way. And last one I'd make there's many more, but I would just stress these ones in the environmental sustainability side of it.
AI data centers use a lot of electricity and water. So that's the downside. But on the positive side, through digital twins modeling the weather and earth sciences, through the amazing efficiencies that AI will achieve across many different industries and through new discoveries in energy sources and other other areas, I think AI will probably be a plus overall, but we still need to monitor and where possible, reduce its energy output.
Thanks. Thank you very much. I should have said at the beginning if you could, when you start speaking, you might care to introduce yourself so that the committee and the wider public outside who might be watching would know who you were as well. I do that now. Yeah. Thanks. My name is Sam Dawes. I'm a senior advisor to the Oxford Martin AI Governance Initiative at the University of Oxford and the founding director of multilateral AI, a former senior civil servant in the Foreign Office and Cabinet Office, and before that worked for the UN, including as first officer to UN Secretary General Kofi Annan.
Thank you very much.
I thank you very much for making yourselves available today for our interrogation and , and addressing this subject of how the multilateral institutions might deal with the problems and difficulties of AI. So I remind you and others that this session is being streamed live on Parliament TV. Transcript will be taken and you will be able to correct any minor changes in in that as well. You'll get a copy of the transcript.
And I remind members that if they have relevant interests, they should declare them. And I declare that I am a senior counsellor with the Washington based Cohen Group. Right at the at the very beginning. So if I could if I could ask the first question. As you know, the the committee is looking at the whole multilateral system as it applies to the United Kingdom and hopefully will make recommendations in relation to to the UK's policy in relation to the various multilateral institutions that exist in the world today.
So that is why we are looking at the whole question of artificial intelligence in the context of the the global picture and the way in which multilateral institutions have already or may in the future, look at these things. So so what are the key governance issues related to AI that that multilateral engagement really needs to address? So if I if I could ask Mr. DOS if he if he would like to ask the answer the question first.
Certainly it's a real pleasure and privilege to be here. There's such a broad list of governance challenges. I'll pick out just a few of them. I think the first would be safety . Agreeing common methods for evaluating advanced AI, sharing risk information and responding to serious incidents. And we've seen in the news just the last few days the importance of Agentic AI and frontier capabilities in the cyber domain.
The second would be international security. From cyber attacks to AI enabled biological threats. Military and more. More generally. Third, regulatory interoperability. There's different approaches in the EU, China, US, UK and so on. Those differences are legitimate, but we need to make sure our systems are remain compatible. The next one, and this is often neglected I think, but is increasingly important, is cultural and linguistic diversity, that AI must not be just for greater efficiency, but it also needs to to reflect the wisdom of of humanity as a whole.
Information integrity. So coordinating responses to election interference that AI can supercharge science and public health. There's huge potential as reasoning models develop and through eventually recursive self-improvement, to have a cascade of new discoveries in mathematics, science, and medicine. And I think governance is important to harness that in the most effective way. And last one I'd make there's many more, but I would just stress these ones in the environmental sustainability side of it.
AI data centers use a lot of electricity and water. So that's the downside. But on the positive side, through digital twins modeling the weather and earth sciences, through the amazing efficiencies that AI will achieve across many different industries and through new discoveries in energy sources and other other areas, I think AI will probably be a plus overall, but we still need to monitor and where possible, reduce its energy output.
Thanks. Thank you very much. I should have said at the beginning if you could, when you start speaking, you might care to introduce yourself so that the committee and the wider public outside who might be watching would know who you were as well. I do that now. Yeah. Thanks. My name is Sam Dawes. I'm a senior advisor to the Oxford Martin AI Governance Initiative at the University of Oxford and the founding director of multilateral AI, a former senior civil servant in the Foreign Office and Cabinet Office, and before that worked for the UN, including as first officer to UN Secretary General Kofi Annan.
Thank you very much.
10:39
Speaker 1
-
Speaker 1
Good morning. Good morning all. What a pleasure and a privilege to be here. By way of introduction, my name is Isabella Wilkinson. I am a senior research fellow at Chatham House, an international policy institute where I'm part of our digital society program. So my work focuses on the global governance of emerging technology and disinformation. I'm also a doctoral student at the Blavatnik School of Government in the University of Oxford, where my research looks at how companies work together on AI security.
My research is supervised by former UK cyber chief Kieran Martin and professor and expert in technology governance, Roxana Radu. Now, to answer the first question, if that's okay to dive straight in, I echo Mr. Dawes's diagnosis of some of the key governance challenge many of you would have come confronted with over, especially over the past few months and particularly over the past few days. I think it's helpful for the committee to visualize key governance challenges as falling into three categories.
And what I'll do is walk through each of these categories, and hopefully this will help us structure some of the answers in the session going forward. The first type of governance challenge is dealing with shared risks. As we've seen over the past few weeks, few days, even few hours, risks generated and triggered by AI are cross border. They are inevitably globally shared and their cross domains. They might jump from one part of the economy to the other, from one domain to another.
These include malicious risks. So we've seen concerns about, say, cyber criminals developing a bioweapon. Loss of control risks. So this is when AI models and systems are behaving exactly as they were designed to. But they are, for example, performing in ways which their designers no longer expect them to do. So. So this is called a loss of control risk. And then finally there's a systemic risk. And so these are risks associated with the adoption or diffusion of AI systems that lead to, say, risks like discrimination or labor displacement.
So to recap, there are malicious risks. There are loss of control risks, and there are systemic risks. And we're facing a governance challenge because all of these risks are interlinked. So we have to be very strategic about how we design governance at multiple different levels. This is a multilateral challenge. Now, the second category of governance challenges, which we're confronting at the moment, relates to trust and information. To echo what Mr.
Dawes said, there is a dearth of trusted information about AI, both its risks, its impacts and the opportunities that we may be able to collectively and multilaterally harness. There is an emerging scientific evidence base, but it is by no means globally sufficient. And what we're missing is a testable picture of risks and opportunities. Now the third category is preparedness for disruption. As we've covered briefly, AI risks are inevitably cross border.
They are inevitably cross-domain. And so a key governance challenge that companies and countries will face is how to prepare for any range of AI related and AI driven disruptions. In other words, a key governance challenge is our lack of a crisis toolbox, networks, strategies, resources to deploy in the event of a major crisis, or even a small one, which triggers broader disruption further down the line. In short, we are facing quite a severe state of gridlock in multilateral AI governance due to these interlinked governance challenges.
I would refer the committee to the UN's preliminary International Scientific Report on AI, which platforms global emerging consensus on the state of risk, and links it to many of the different governance challenges that I've mentioned in my remarks.
My research is supervised by former UK cyber chief Kieran Martin and professor and expert in technology governance, Roxana Radu. Now, to answer the first question, if that's okay to dive straight in, I echo Mr. Dawes's diagnosis of some of the key governance challenge many of you would have come confronted with over, especially over the past few months and particularly over the past few days. I think it's helpful for the committee to visualize key governance challenges as falling into three categories.
And what I'll do is walk through each of these categories, and hopefully this will help us structure some of the answers in the session going forward. The first type of governance challenge is dealing with shared risks. As we've seen over the past few weeks, few days, even few hours, risks generated and triggered by AI are cross border. They are inevitably globally shared and their cross domains. They might jump from one part of the economy to the other, from one domain to another.
These include malicious risks. So we've seen concerns about, say, cyber criminals developing a bioweapon. Loss of control risks. So this is when AI models and systems are behaving exactly as they were designed to. But they are, for example, performing in ways which their designers no longer expect them to do. So. So this is called a loss of control risk. And then finally there's a systemic risk. And so these are risks associated with the adoption or diffusion of AI systems that lead to, say, risks like discrimination or labor displacement.
So to recap, there are malicious risks. There are loss of control risks, and there are systemic risks. And we're facing a governance challenge because all of these risks are interlinked. So we have to be very strategic about how we design governance at multiple different levels. This is a multilateral challenge. Now, the second category of governance challenges, which we're confronting at the moment, relates to trust and information. To echo what Mr.
Dawes said, there is a dearth of trusted information about AI, both its risks, its impacts and the opportunities that we may be able to collectively and multilaterally harness. There is an emerging scientific evidence base, but it is by no means globally sufficient. And what we're missing is a testable picture of risks and opportunities. Now the third category is preparedness for disruption. As we've covered briefly, AI risks are inevitably cross border.
They are inevitably cross-domain. And so a key governance challenge that companies and countries will face is how to prepare for any range of AI related and AI driven disruptions. In other words, a key governance challenge is our lack of a crisis toolbox, networks, strategies, resources to deploy in the event of a major crisis, or even a small one, which triggers broader disruption further down the line. In short, we are facing quite a severe state of gridlock in multilateral AI governance due to these interlinked governance challenges.
I would refer the committee to the UN's preliminary International Scientific Report on AI, which platforms global emerging consensus on the state of risk, and links it to many of the different governance challenges that I've mentioned in my remarks.
10:43
10:43
Lord Ahmad of Wimbledon
-
-
Lord Ahmad of Wimbledon
Thank you, Lord chairman, and welcome to you both. And it's good to see you again, Sam. We see a variety of international organizations. You mentioned one of them, the United Nations, but also the G7, OECD, the Council of Europe, all looking at this whole issue of common international governance standards from our own experiences. Previously, when I served in government, looking at cyber governance, there was always a challenge the Chinese version, the American version, the European model, which would apply.
How do you characterize where we are in terms of the current state of play in different countries? And also, I suppose, the appetite that there is for internet governance? Isabella, you touched on the fact also that there are initiatives underway, the UN and whether that's effective. But each country in a way is pushing its own agenda. So in that context, perhaps I can start with your good self. As to building on what you've already said, is there space for one organization or indeed a series of organizations to follow a particular governance model?
How do you characterize where we are in terms of the current state of play in different countries? And also, I suppose, the appetite that there is for internet governance? Isabella, you touched on the fact also that there are initiatives underway, the UN and whether that's effective. But each country in a way is pushing its own agenda. So in that context, perhaps I can start with your good self. As to building on what you've already said, is there space for one organization or indeed a series of organizations to follow a particular governance model?
10:44
Speaker 1
-
Speaker 1
Thank you, Lord Ahmed. I broadly agree with your assessment. We are seeing a multitude of different governance initiatives at different levels. So different regional groupings with different aims, with different modalities. It's helpful to characterize this as a sort of global patchwork currently stuck in a state, as I mentioned before, of intense gridlock. And there are many reasons for this gridlock. Some of the governance challenges I set out in my first answer, some of the competition dynamics between the US and China that you all will be incredibly familiar with.
But the bottom line here is that multilateral AI governance as pushed forward in a variety of different regional groupings, international institutions, other arrangements, is nascent. It's still taking its first steps. There is much dialogue, there's much talk. But as it stands, our case studies of implementation and of actioning these commitments and these regulations are just years old. So we're talking about an incredibly young space. A key example of this is the UN's inaugural meeting on AI governance in July.
They convened hundreds, I think, over 3000 participants and over 100 UN member states to discuss ways forward for AI governance. It was a valuable platform. Some people called it a dialogue of dialogues, but it didn't lead to a binding commitment. And so what we're facing right now in the state of play are a variety of different spaces for dialogue and for consensus building, but lack of connective tissue to actionable implementation. But this isn't an impossible bridge to cross.
And I look forward to exchanging ideas on how to do so. I'm happy to walk through some areas of success, of failure, of inertia, if that would be helpful. Or I can hand over to Mr. Dawes. Great. I'll go ahead and please cut me off. If I'm taking too long within this global patchwork, this somewhat severe gridlock that we've sketched out, there are some spotlights of success. I would categorize these as successes and evidence and documentation as a basis for governance.
From a technical perspective, it's really exciting to see emergence of global best practice on how AI incidences are reported on. So the way that we report on what goes wrong and how we mitigate it is becoming increasingly standardized. And as we've seen in other areas of governance, this is an essential risk mitigation measure. An example of this is, as I mentioned before, the UN International Scientific Report and the international AI Safety Report, which kicked off after the 2023 Bletchley Summit.
Another example is the standardization of approaches to transparency around AI, which the G7 and OECD have very much been leading over the past few years, and what we see here as a model of success is a sort of theory of change. Small groupings come together and agree that a technical best practice in governance works, and increasingly it sort of snowballs and scales to become multilateral best practice. We've seen this success in evidence, in documentation, although it is still nascent, and we should learn from this impact pathway for the areas of failure and inertia in multilateral AI governance, which I'm going to cover now.
Nothing about pacing the frontier, which is terminology you would have all encountered over the past few days, is internationally mandated. The bottom line here is that there are some binding rules. For example, the EU AI act has reporting requirements on AI models and providers, but there's nothing internationally mandated. This is a massive governance gap, and I'm sure that we'll discuss this later in the session. There are areas of inertia in multilateral AI governance where, for example, there's growing global consensus that something should be a priority, even growing global outcry, but a lack of action due to this gridlock.
A really chilling example of this are the human rights abuses that AI systems may enable the integration of AI in a military domain, leading to severe civilian harms in places ranging from Gaza to Iran. Now, it isn't to say that there hasn't been movement. Again, on a smaller level, the Council of Europe Convention on AI puts human rights front and center. But returning to our theory of change, there needs to be connective tissue between consensus and shared language in smaller groupings and multilateral consensus.
Let me finish my remarks by underlining the fact that we are in a moment of intense crisis and turbulence in multilateral AI governance. But as we've learned, and I'm sure this committee has learned from many other areas of governance in the past, these conditions create a window of opportunity for dealing with the gridlock. And I'd love us to return to this idea later in the session.
But the bottom line here is that multilateral AI governance as pushed forward in a variety of different regional groupings, international institutions, other arrangements, is nascent. It's still taking its first steps. There is much dialogue, there's much talk. But as it stands, our case studies of implementation and of actioning these commitments and these regulations are just years old. So we're talking about an incredibly young space. A key example of this is the UN's inaugural meeting on AI governance in July.
They convened hundreds, I think, over 3000 participants and over 100 UN member states to discuss ways forward for AI governance. It was a valuable platform. Some people called it a dialogue of dialogues, but it didn't lead to a binding commitment. And so what we're facing right now in the state of play are a variety of different spaces for dialogue and for consensus building, but lack of connective tissue to actionable implementation. But this isn't an impossible bridge to cross.
And I look forward to exchanging ideas on how to do so. I'm happy to walk through some areas of success, of failure, of inertia, if that would be helpful. Or I can hand over to Mr. Dawes. Great. I'll go ahead and please cut me off. If I'm taking too long within this global patchwork, this somewhat severe gridlock that we've sketched out, there are some spotlights of success. I would categorize these as successes and evidence and documentation as a basis for governance.
From a technical perspective, it's really exciting to see emergence of global best practice on how AI incidences are reported on. So the way that we report on what goes wrong and how we mitigate it is becoming increasingly standardized. And as we've seen in other areas of governance, this is an essential risk mitigation measure. An example of this is, as I mentioned before, the UN International Scientific Report and the international AI Safety Report, which kicked off after the 2023 Bletchley Summit.
Another example is the standardization of approaches to transparency around AI, which the G7 and OECD have very much been leading over the past few years, and what we see here as a model of success is a sort of theory of change. Small groupings come together and agree that a technical best practice in governance works, and increasingly it sort of snowballs and scales to become multilateral best practice. We've seen this success in evidence, in documentation, although it is still nascent, and we should learn from this impact pathway for the areas of failure and inertia in multilateral AI governance, which I'm going to cover now.
Nothing about pacing the frontier, which is terminology you would have all encountered over the past few days, is internationally mandated. The bottom line here is that there are some binding rules. For example, the EU AI act has reporting requirements on AI models and providers, but there's nothing internationally mandated. This is a massive governance gap, and I'm sure that we'll discuss this later in the session. There are areas of inertia in multilateral AI governance where, for example, there's growing global consensus that something should be a priority, even growing global outcry, but a lack of action due to this gridlock.
A really chilling example of this are the human rights abuses that AI systems may enable the integration of AI in a military domain, leading to severe civilian harms in places ranging from Gaza to Iran. Now, it isn't to say that there hasn't been movement. Again, on a smaller level, the Council of Europe Convention on AI puts human rights front and center. But returning to our theory of change, there needs to be connective tissue between consensus and shared language in smaller groupings and multilateral consensus.
Let me finish my remarks by underlining the fact that we are in a moment of intense crisis and turbulence in multilateral AI governance. But as we've learned, and I'm sure this committee has learned from many other areas of governance in the past, these conditions create a window of opportunity for dealing with the gridlock. And I'd love us to return to this idea later in the session.
10:49
Lord Ahmad of Wimbledon
-
-
Lord Ahmad of Wimbledon
Thank you. So I'm just picking on a couple of those. There's a general point that the whole international rules based order is in is chaotic. Is there an appetite to use multilateral governance structures to pursue this? Everyone believes something should be done, but you have others who would say, yes, it should be done according to our way, and others who are not quite clear which method of governance to follow.
10:50
Lord Robertson of Port Ellen
-
-
Lord Robertson of Port Ellen
Yes, I think there's a distinction . Governance and regulation between voluntary principles and enforcement. And I agree with Miss Wilkinson that it's both fragmentary and patchwork is what we see. I've actually been quite surprised by the degree of universal through the UN regional and many lateral initiatives that there have been over the last few years. The UN ones, I think, are in part to the fact that Antonio Guterres was the first unsg with a science and engineering background.
So he set up a UN advisory board. And then, as we've heard, we've had a scientific panel and a largely talk shop focused global global dialogue, but valuable in itself. I would probably divide it across those three things we've seen, certainly under the previous US administration, a willingness to set principles at the UN through two UN General Assembly resolutions in lockstep with the Chinese. So there was a resolution first on 2004 on on AI safety and then one on AI capacity building.
The US put forward the first backed by China. China put forward the second backed by the US. So I think there was a real recognition early on and a desire between those two countries to to have some kind of framework, the regional level. The EU has had very much regulation, risk based, law based approach, while other regions have focused more on AI sovereignty. In the case of Africa, focused on data sets, privacy languages, Asean, similar, Central Asia, similar Latin America.
I think the most interesting area in answer to your direct question is, is the innovative unilateral initiatives that have taken place there. You've got the kind of Western based ones. The G7 has been mentioned, the OECD, which expanded, adding in Brazil and India through its integrating the GP Global partnership on AI, the Bletchley process that Prime Minister Rishi Sunak initiated was an important mini lateral, which has continued through different incarnations in different countries.
Seoul, Paris, New Delhi next year will be . Singapore has been quite important , and the Singapore consensus, which is a research consensus bringing together US, Chinese, UK and other researchers from 13 countries, focusing in the last incarnation on the risks from a genetic AI. So you've got these Western based mini laterals, these kind of neutral mini laterals, as it were, crossing over the the Digital Cooperation Organization, an initiative from the Kingdom of Saudi Arabia.
I think it's an important one that crosses different regions. The Netherlands and the Republic of Korea have a mini lateral on responsible AI in the in the military realm. And then you've got the more Chinese centered ones. The Shanghai Cooperation Organization's AI engagement, BRICs plus, the Digital Belt and Road Initiative, and most recently, I was out in Shanghai in July for it, the launch of the world AI Cooperation Organization, which builds on a real Chinese focus on basically the diffusion of AI technologies, not necessarily at the frontier.
So I think there is a real mixture of of different approaches from the universe or the regional to the to the mini lateral. I agree, I think we do need connective tissue, but between them and we need also to embrace the private sector because they are key to solving this. So there's things like the Frontier Model forum, the private sector's involvement in the various safety summits. In terms of reporting, we've got frameworks like the G20 and APEC, which could also be a good a good places to to embrace private sector involvement.
But that's very much in its infancy in that area.
So he set up a UN advisory board. And then, as we've heard, we've had a scientific panel and a largely talk shop focused global global dialogue, but valuable in itself. I would probably divide it across those three things we've seen, certainly under the previous US administration, a willingness to set principles at the UN through two UN General Assembly resolutions in lockstep with the Chinese. So there was a resolution first on 2004 on on AI safety and then one on AI capacity building.
The US put forward the first backed by China. China put forward the second backed by the US. So I think there was a real recognition early on and a desire between those two countries to to have some kind of framework, the regional level. The EU has had very much regulation, risk based, law based approach, while other regions have focused more on AI sovereignty. In the case of Africa, focused on data sets, privacy languages, Asean, similar, Central Asia, similar Latin America.
I think the most interesting area in answer to your direct question is, is the innovative unilateral initiatives that have taken place there. You've got the kind of Western based ones. The G7 has been mentioned, the OECD, which expanded, adding in Brazil and India through its integrating the GP Global partnership on AI, the Bletchley process that Prime Minister Rishi Sunak initiated was an important mini lateral, which has continued through different incarnations in different countries.
Seoul, Paris, New Delhi next year will be . Singapore has been quite important , and the Singapore consensus, which is a research consensus bringing together US, Chinese, UK and other researchers from 13 countries, focusing in the last incarnation on the risks from a genetic AI. So you've got these Western based mini laterals, these kind of neutral mini laterals, as it were, crossing over the the Digital Cooperation Organization, an initiative from the Kingdom of Saudi Arabia.
I think it's an important one that crosses different regions. The Netherlands and the Republic of Korea have a mini lateral on responsible AI in the in the military realm. And then you've got the more Chinese centered ones. The Shanghai Cooperation Organization's AI engagement, BRICs plus, the Digital Belt and Road Initiative, and most recently, I was out in Shanghai in July for it, the launch of the world AI Cooperation Organization, which builds on a real Chinese focus on basically the diffusion of AI technologies, not necessarily at the frontier.
So I think there is a real mixture of of different approaches from the universe or the regional to the to the mini lateral. I agree, I think we do need connective tissue, but between them and we need also to embrace the private sector because they are key to solving this. So there's things like the Frontier Model forum, the private sector's involvement in the various safety summits. In terms of reporting, we've got frameworks like the G20 and APEC, which could also be a good a good places to to embrace private sector involvement.
But that's very much in its infancy in that area.
10:54
10:54
Baroness Crawley
-
-
Baroness Crawley
Thank you, chair, and you're very welcome. And it's good to hear that there is this myriad of connectivity, even if it's not strongly underpinned by law at the moment. However, what are the challenges? Do you see the really hard core challenges to multilateral cooperation on issues such as artificial intelligence and common international standards? We were given a very useful diagram in our briefing, which was the stack, and it's on page eight.
And and it it has in it the key governance bodies for each layer of the stack. So the social layer, ChatGPT, Gemini, etc., the logical layer, OpenAI, DeepMind and the infrastructure layer, Amazon, Microsoft, Google, etc.. So the challenges to agreeing standards, are they more in the in the top layer than in the technical layer? Are they more in the logical layer than in the human rights layer? What? Where are they? Where are the challenges?
And and it it has in it the key governance bodies for each layer of the stack. So the social layer, ChatGPT, Gemini, etc., the logical layer, OpenAI, DeepMind and the infrastructure layer, Amazon, Microsoft, Google, etc.. So the challenges to agreeing standards, are they more in the in the top layer than in the technical layer? Are they more in the logical layer than in the human rights layer? What? Where are they? Where are the challenges?
10:56
Lord Robertson of Port Ellen
-
-
Lord Robertson of Port Ellen
Yeah, I start and. I think I would begin by saying that there are three overarching challenges that, that pertain to the whole stack. The first is the geopolitical and dual use nature of, of AI. The, the frontier models that we're talking about have civilian uses, but they can also be used in military contexts. If you look at nearly all of the international agreements, the UN, OECD Council of Europe and so on, they all have a sentence saying this applies just to the civilian use of AI.
So that is one of the fundamental things. And feeding into that the dual use nature, we've got the geopolitical competition between the US and China, where the US has imposed export controls on the export of high end chips and also lithography, EUV lithography equipment to etch the chips. In conjunction with the Netherlands. So there's a lot of tension around around the geopolitical space. And there's also different there's a diffusion competition between the US and China.
China has taken the approach of offering a full stack, low cost. Interoperable approach, which is focused very much on what they call being tuned to context, culture and language so that it might be tuned for Southeast Asian education or African agriculture. Or. So there's a real attempt by China to get effectively market share and also be helpful to the developing world and helpful to China in setting its standards as the global standards through diffusion.
Its approach has also been focused on integrating AI not at the frontier, but integrating AI through embodied AI, through manufacturing robotics into every different industry. The US approach that the scaling approach has been more focused on the aspirations for artificial general intelligence and and models that can basically do everything that they've got capacities in robotics, they've got capacities in brain, computer interface and so on. So it's not an either or, but there's been there's different dynamics.
And I think the Trump administration woke up to the diffusion approach of China and therefore took some steps to move away from Biden's restrictions on exporting high end chips to countries like the UAE and Saudi Arabia because of a recognition that competition. So that's the first level. The second is commercial sensitivity. Governance is so difficult because individual companies, whether they're American or Chinese or British companies, are unwilling to reveal their model weights or the data sets used to train their models because it could undermine their profitability.
And the third is the nature of the technology itself. It's opaque and it's fast moving. Some of the the best scientists, technologists I speak to inside the companies that are developing these, these models don't really understand how the models come to their conclusions, how what happens within the black box. So to govern or regulate that is therefore inherently difficult. And we're seeing the changing nature of the technology with with Agentic AI.
It's not just a question of how do you test, evaluate, secure, ensure that a model is safe before it's released. That's super important, but that's one part of the puzzle with Agentic capabilities. Models now do things and they interact in the world. Agents work with each other in unexpected ways. They they collude with each other or cooperate with each other to obtain their their instructed end result.
And we're seeing emergent qualities, for example, of deception of agents tricking humans in order not to be switched off or tricking humans in order to fulfill their assigned purpose. So it's a misalignment problem. So we, we, we need also to govern AI in full life cycle. So after its release, see how it develops in the wild, as it were, because we're going to have 1000 million agents interacting in the wild and bad actors, whether they're terrorist group or organized crime, are also harness some of these tools.
Or it could be another it could be a bad state actor. So there are multiple challenges with governance and regulation.
So that is one of the fundamental things. And feeding into that the dual use nature, we've got the geopolitical competition between the US and China, where the US has imposed export controls on the export of high end chips and also lithography, EUV lithography equipment to etch the chips. In conjunction with the Netherlands. So there's a lot of tension around around the geopolitical space. And there's also different there's a diffusion competition between the US and China.
China has taken the approach of offering a full stack, low cost. Interoperable approach, which is focused very much on what they call being tuned to context, culture and language so that it might be tuned for Southeast Asian education or African agriculture. Or. So there's a real attempt by China to get effectively market share and also be helpful to the developing world and helpful to China in setting its standards as the global standards through diffusion.
Its approach has also been focused on integrating AI not at the frontier, but integrating AI through embodied AI, through manufacturing robotics into every different industry. The US approach that the scaling approach has been more focused on the aspirations for artificial general intelligence and and models that can basically do everything that they've got capacities in robotics, they've got capacities in brain, computer interface and so on. So it's not an either or, but there's been there's different dynamics.
And I think the Trump administration woke up to the diffusion approach of China and therefore took some steps to move away from Biden's restrictions on exporting high end chips to countries like the UAE and Saudi Arabia because of a recognition that competition. So that's the first level. The second is commercial sensitivity. Governance is so difficult because individual companies, whether they're American or Chinese or British companies, are unwilling to reveal their model weights or the data sets used to train their models because it could undermine their profitability.
And the third is the nature of the technology itself. It's opaque and it's fast moving. Some of the the best scientists, technologists I speak to inside the companies that are developing these, these models don't really understand how the models come to their conclusions, how what happens within the black box. So to govern or regulate that is therefore inherently difficult. And we're seeing the changing nature of the technology with with Agentic AI.
It's not just a question of how do you test, evaluate, secure, ensure that a model is safe before it's released. That's super important, but that's one part of the puzzle with Agentic capabilities. Models now do things and they interact in the world. Agents work with each other in unexpected ways. They they collude with each other or cooperate with each other to obtain their their instructed end result.
And we're seeing emergent qualities, for example, of deception of agents tricking humans in order not to be switched off or tricking humans in order to fulfill their assigned purpose. So it's a misalignment problem. So we, we, we need also to govern AI in full life cycle. So after its release, see how it develops in the wild, as it were, because we're going to have 1000 million agents interacting in the wild and bad actors, whether they're terrorist group or organized crime, are also harness some of these tools.
Or it could be another it could be a bad state actor. So there are multiple challenges with governance and regulation.
11:01
11:01
11:01
Lord Grocott
-
-
Lord Grocott
Yes, sir. With your with your perspective of your background and experience, would you like to reflect for a moment or two for for our benefit on some of the language that's being used about AI at the moment from, you know, I don't know, wiping out the human race within a measurable period of time. Another thing that springs to mind, I was on a committee looking at AI in defence, and the phrase killer robots came up from time to time in the evidence.
I mean, where are you on that spectrum? I'll make it a specific as this. Should we be really worried, mildly worried or at the stage of sort of academic reflection? And if it's really worried, how imminent is it?
I mean, where are you on that spectrum? I'll make it a specific as this. Should we be really worried, mildly worried or at the stage of sort of academic reflection? And if it's really worried, how imminent is it?
11:02
Speaker 1
-
Speaker 1
I'm happy to jump in. In my estimation, the hyper militarization and anthropomorphic anthropomorphization. Sorry, that word is a bit of a mouthful is incredibly unhelpful and inappropriate to the moment we find ourselves in relying on this sort of language, on this sort of discourse not only obfuscates governance solutions, but it has a polarising impact. And let's walk through some of these impacts and consider different framings, which may help us reimagine the moment we're in and try to unblock some of the gridlock that we've diagnosed currently in the session today , over militarization feeds into zero sum delusions.
Now, we've seen the regurgitation of Cold War and arms race framings around the US-China race for technological supremacy. And indeed, it may be a race for technological supremacy, but it assumes that attaining artificial general intelligence or artificial superintelligence, which for the most part is an unmeasured benchmark, means the game is up and there's nothing else to do. It's much more appropriate to reframe to words that we've discussed in the session today a patchwork, the existence of multiple paradigms and multiple pathways pushing back against the assumptions that some sort of technological supremacy and end point is inevitable, that we are on a sort of determined or predetermined pathway to that point.
It's inappropriate also, because it leads us into the assumption that the world is divided into Cold War esque blocks the US versus China, US technology and infrastructure versus Chinese technology and infrastructure. Again, it makes sense that people tend to divide into these sorts of blocks when they're talking about how to govern, develop, deploy, diffuse. But doing so obfuscates the existence and really the reality of alternative and third pathways, the development, say, of open source models in the global majority UK and European aspirations for building and developing technology in the national interest.
And so this leads us into a sort of powerful reimagining, pushing us away from catastrophizing, anthropomorphizing, militarizing, and towards something which is rooted in openness in plurality. It may be more helpful to reimagine the development of AI capabilities as a sort of moonshot. It may be helpful to bake in the acceptance of plurality to conversations about governance, again, pushing back against an inevitability narrative and reclaiming some agency. Because, as we've seen over the past week, reclaiming agency and sobriety in discourse is as important as it in discourse as it is in the development and deployment of the technology itself.
And I'd refer the committee to a really interesting anthology on reimagining the AI arms race, which pitches a variety of different framings.
Now, we've seen the regurgitation of Cold War and arms race framings around the US-China race for technological supremacy. And indeed, it may be a race for technological supremacy, but it assumes that attaining artificial general intelligence or artificial superintelligence, which for the most part is an unmeasured benchmark, means the game is up and there's nothing else to do. It's much more appropriate to reframe to words that we've discussed in the session today a patchwork, the existence of multiple paradigms and multiple pathways pushing back against the assumptions that some sort of technological supremacy and end point is inevitable, that we are on a sort of determined or predetermined pathway to that point.
It's inappropriate also, because it leads us into the assumption that the world is divided into Cold War esque blocks the US versus China, US technology and infrastructure versus Chinese technology and infrastructure. Again, it makes sense that people tend to divide into these sorts of blocks when they're talking about how to govern, develop, deploy, diffuse. But doing so obfuscates the existence and really the reality of alternative and third pathways, the development, say, of open source models in the global majority UK and European aspirations for building and developing technology in the national interest.
And so this leads us into a sort of powerful reimagining, pushing us away from catastrophizing, anthropomorphizing, militarizing, and towards something which is rooted in openness in plurality. It may be more helpful to reimagine the development of AI capabilities as a sort of moonshot. It may be helpful to bake in the acceptance of plurality to conversations about governance, again, pushing back against an inevitability narrative and reclaiming some agency. Because, as we've seen over the past week, reclaiming agency and sobriety in discourse is as important as it in discourse as it is in the development and deployment of the technology itself.
And I'd refer the committee to a really interesting anthology on reimagining the AI arms race, which pitches a variety of different framings.
11:05
Lord Robertson of Port Ellen
-
-
Lord Robertson of Port Ellen
Thank you very much, Mr. Dose. Do you want to answer the question? I could add just a few. That's a fantastic answer. I I'm extremely worried, I would say, but I'm also optimistic. I think the risks are very real of the ability to for bad actors to develop. Bioweapons, for example, is one of the scariest examples. The loss of control issue of AI misalignment is very genuine and we're seeing evidence of it.
But there is a genuine willingness to to address it across different stakeholders. So I think , I think that there are solutions, and I agree also on the on the narrative, I think that often Cold War narratives are only go so far when it comes to AI. It's very it's a very different phenomena than nuclear. It's something where actually, if the risks are downstream, it means that every country in the world needs to be engaged if we're going to solve it.
It's a bit like terrorism. You can't have a space in, you know, in Somalia, Sudan, Yemen, etc., where terrorists were able to, to , to regroup. You need every, every government to be empowered around the world to, to, to address the, the possible fallout from, from, from AI. And I think that the, the kind of narrative around Cold war and arms race can be self-fulfilling. It can encourage secrecy, it can weaken safeguards.
It can actually help create the risk that it that it predicts. So a better framing is, I think, strategic competition with with, with shared security obligations would be my approach. And I think there's some great examples from the Cold War. And since of what we've done when there isn't trust between parties. So confidence building measures developed in the kind of arms control , cyber, bio chemical, nuclear weapons areas.
The UN Institute for Disarmament Research has been doing some good work on on which of those actually is transferable in the AI domain and which is which is not. So I think confidence building measures, kind of red telephones, as it were, is good. And there was an encouraging part of President XI Jinping's speech at the world AI conference, where it's the first time that a leader of a major country has articulated the need for prevention, for addressing loss of control and for emergency management cooperation.
So I think that we're seeing some signs that President Trump, in his last UN General Assembly speech, talked about the need for international cooperation on the nexus of AI and and bio risks. So I think we're seeing an agenda where actually there's willingness as long as the concerns of all sides, including that, that this doesn't impede innovation, are respected.
But there is a genuine willingness to to address it across different stakeholders. So I think , I think that there are solutions, and I agree also on the on the narrative, I think that often Cold War narratives are only go so far when it comes to AI. It's very it's a very different phenomena than nuclear. It's something where actually, if the risks are downstream, it means that every country in the world needs to be engaged if we're going to solve it.
It's a bit like terrorism. You can't have a space in, you know, in Somalia, Sudan, Yemen, etc., where terrorists were able to, to , to regroup. You need every, every government to be empowered around the world to, to, to address the, the possible fallout from, from, from AI. And I think that the, the kind of narrative around Cold war and arms race can be self-fulfilling. It can encourage secrecy, it can weaken safeguards.
It can actually help create the risk that it that it predicts. So a better framing is, I think, strategic competition with with, with shared security obligations would be my approach. And I think there's some great examples from the Cold War. And since of what we've done when there isn't trust between parties. So confidence building measures developed in the kind of arms control , cyber, bio chemical, nuclear weapons areas.
The UN Institute for Disarmament Research has been doing some good work on on which of those actually is transferable in the AI domain and which is which is not. So I think confidence building measures, kind of red telephones, as it were, is good. And there was an encouraging part of President XI Jinping's speech at the world AI conference, where it's the first time that a leader of a major country has articulated the need for prevention, for addressing loss of control and for emergency management cooperation.
So I think that we're seeing some signs that President Trump, in his last UN General Assembly speech, talked about the need for international cooperation on the nexus of AI and and bio risks. So I think we're seeing an agenda where actually there's willingness as long as the concerns of all sides, including that, that this doesn't impede innovation, are respected.
11:09
11:09
11:09
Lord Robertson of Port Ellen
-
-
Lord Robertson of Port Ellen
A great question. And I think nothing is inevitable. And what we've seen with the development of AI is it's jagged, like some models. They can't tell you how many hours are in the word raspberry, but they can solve Olympiad level Olympic problems. And so I think it's really hard to predict, actually, whether I personally doubt that we can ever say AI has reached artificial general intelligence or artificial superintelligence. I don't think there'll be a moment when we know that one country has complete dominance in all domains security, economic, etc., over another.
And I think that's that uncertainty is one of the reasons why these analogies around arms control and winner takes all break down. So worst case scenario is pretty grim. But I think there's plenty of other challenges that we know that are difficult, wicked problems like climate change that that actually the world is grappling with and we may or may not solve them. AI is not top of my list in terms of that, I think existential risk, but I think it's something that we need to take seriously.
But, you know, the, the, the whole nuclear weapons side of thing is surrounded by mystique and secrecy and protocols, you know, and which and also sort of costs factors as well, which have given the international organization some degree of traction, but that doesn't apply for chemical and biological weapons, where there is largely an unregulated space. So why should you be optimistic that you can control artificial intelligence in a way that we have not even been able to do with chemical and biological in? Very difficult, I think, partly because there's a huge amount of economic value committed to the success of AI.
And I think that companies, as we've seen in the last few days, are waking up to the fact that safety is an important. It is essential if they're not going to lose the economic value that they that they put into it. If public loses trust in AI, governments lose trust. Parliaments lose trust, then then that's, you know, very difficult. So I think that that's the incentives are better aligned. I think you make a great point about nuclear, chemical and biological weapons.
I think there is a disparity, actually , between the nuclear and chemical regime, where you have the Non-Proliferation treaty for nuclear and the organisation for the Prohibition of Chemical Weapons or Chemical, and they are large organisations with large secretariats. The Biological Weapons Convention, by contrast, has only four full time staff members and ten kind of secondments so tiny in comparison. And there was. So there's a real need. And I think the UK, because its leadership, both in life sciences and in with Portland and, you know, the sort of looking at the risks from bio biosecurity and biosafety has has a real contribution to make here in developing the support for the biological weapons convention, specifically because AI superpower is an AI.
Both increases the positive benefits in the life sciences, but also makes it far more likely that biological weapons may be produced by bad actors. So I think there's interesting we could talk later on it, perhaps, but there's some interesting proposals, for example, from Kazakhstan on creating an international agency for biosecurity. And there's an existing work being done to set up a scientific committee within the BWC. So yeah. So I think all of those are important, I think AI solutions are the incentives.
Are there a Baroness pressure?
And I think that's that uncertainty is one of the reasons why these analogies around arms control and winner takes all break down. So worst case scenario is pretty grim. But I think there's plenty of other challenges that we know that are difficult, wicked problems like climate change that that actually the world is grappling with and we may or may not solve them. AI is not top of my list in terms of that, I think existential risk, but I think it's something that we need to take seriously.
But, you know, the, the, the whole nuclear weapons side of thing is surrounded by mystique and secrecy and protocols, you know, and which and also sort of costs factors as well, which have given the international organization some degree of traction, but that doesn't apply for chemical and biological weapons, where there is largely an unregulated space. So why should you be optimistic that you can control artificial intelligence in a way that we have not even been able to do with chemical and biological in? Very difficult, I think, partly because there's a huge amount of economic value committed to the success of AI.
And I think that companies, as we've seen in the last few days, are waking up to the fact that safety is an important. It is essential if they're not going to lose the economic value that they that they put into it. If public loses trust in AI, governments lose trust. Parliaments lose trust, then then that's, you know, very difficult. So I think that that's the incentives are better aligned. I think you make a great point about nuclear, chemical and biological weapons.
I think there is a disparity, actually , between the nuclear and chemical regime, where you have the Non-Proliferation treaty for nuclear and the organisation for the Prohibition of Chemical Weapons or Chemical, and they are large organisations with large secretariats. The Biological Weapons Convention, by contrast, has only four full time staff members and ten kind of secondments so tiny in comparison. And there was. So there's a real need. And I think the UK, because its leadership, both in life sciences and in with Portland and, you know, the sort of looking at the risks from bio biosecurity and biosafety has has a real contribution to make here in developing the support for the biological weapons convention, specifically because AI superpower is an AI.
Both increases the positive benefits in the life sciences, but also makes it far more likely that biological weapons may be produced by bad actors. So I think there's interesting we could talk later on it, perhaps, but there's some interesting proposals, for example, from Kazakhstan on creating an international agency for biosecurity. And there's an existing work being done to set up a scientific committee within the BWC. So yeah. So I think all of those are important, I think AI solutions are the incentives.
Are there a Baroness pressure?
11:13
Baroness Prashar
-
-
Baroness Prashar
Thank you chair. Thank you so much for your very helpful answers so far. My question is specifically about the UK, because we've heard from a number of witnesses that UK possesses the necessary capability to effectively shape the multilateral conversations on the future of AI. Do you agree with that assessment? And if you do, what are the core capabilities and attributes that UK brings to the table, and is the government making the best use of it? And if not, how can they make the best use of it?
11:14
Speaker 1
-
Speaker 1
Thank you. I have full confidence in the UK's capacity and agency to shape multilateral AI governance, but the UK government and its supportive ecosystem must tread carefully and act strategically and use this window of opportunity, of visibility of salience to advance agreement and consensus on priority areas, which I anticipate we'll get to later in the session. Now, in my approximation, the UK has three US peas. It has three unique selling points when it comes to its engagement on AI multilateralism.
The first is its world leading technical expertise. Now the UK's AI Security Institute and its supportive research and innovation ecosystem is undoubtedly world leading, and it has a high level of trust and authority. This is a huge part of the UK's offering to potential partners and allies when it comes to AI multilateralism, we can't forget the supportive ecosystem that enables the curation and platforming of this technical expertise. I'm thinking here of innovation hubs and startups, and Oxford and Cambridge of the Advanced Research and Innovation Agency of UK HRI as well.
Now the second component of the UK's USP on AI multilateralism is its diplomatic heritage. I'm pleased to hear that Bletchley has come up time and time again in our session today, bringing together representatives from the US and China to talk about AI, security and safety is a massive win and the UK has continued to engage. As Mr. Dawes has pointed out in innovative unilateralism in the years that followed the third area where the UK has.
I would argue, a unique selling point is. In its partnerships. The UK boasts a strong alliance with the US on AI and a degree of alignment in. You know, the way. For example, aspects of models are tested. Although there is a little bit of controversy here with Anthropic's decision not to let the AI Security Institute test its latest model, which I would be pleased to discuss later in the session.
But nonetheless, the UK US Partnership on Technology and AI, in addition to the UK's role as a proactive middle power, engaged in innovative mini naturalism. Working as a sort of bridge with Europe is a key strength. Mr. Dawes.
The first is its world leading technical expertise. Now the UK's AI Security Institute and its supportive research and innovation ecosystem is undoubtedly world leading, and it has a high level of trust and authority. This is a huge part of the UK's offering to potential partners and allies when it comes to AI multilateralism, we can't forget the supportive ecosystem that enables the curation and platforming of this technical expertise. I'm thinking here of innovation hubs and startups, and Oxford and Cambridge of the Advanced Research and Innovation Agency of UK HRI as well.
Now the second component of the UK's USP on AI multilateralism is its diplomatic heritage. I'm pleased to hear that Bletchley has come up time and time again in our session today, bringing together representatives from the US and China to talk about AI, security and safety is a massive win and the UK has continued to engage. As Mr. Dawes has pointed out in innovative unilateralism in the years that followed the third area where the UK has.
I would argue, a unique selling point is. In its partnerships. The UK boasts a strong alliance with the US on AI and a degree of alignment in. You know, the way. For example, aspects of models are tested. Although there is a little bit of controversy here with Anthropic's decision not to let the AI Security Institute test its latest model, which I would be pleased to discuss later in the session.
But nonetheless, the UK US Partnership on Technology and AI, in addition to the UK's role as a proactive middle power, engaged in innovative mini naturalism. Working as a sort of bridge with Europe is a key strength. Mr. Dawes.
11:16
Lord Robertson of Port Ellen
-
-
Lord Robertson of Port Ellen
Yeah, I agree with all of that. I would add, I think there's real examples where the UK has led the world in science diplomacy. An example was the acculturation of antimicrobial resistance was it was a double act between chief medical officer Sally Davies, Dame Sally Davies, the time and and ambassador now Dame Karen Pierce in Geneva when she was ambassador there. And they did the real legwork of going regionally to Africa to to Asia Pacific, to Latin America and so on, to getting agreements in veterinary science, in, in animal health, in agriculture, through the different UN bodies, and then taking it to the UN General Assembly.
So we've got real experience of how you actually. Get buy in from around the world for crucial issues. We've got the huge expertise in AI safety from the AI Security Institute and , and so on. And biosafety, as I mentioned before, I think is an area where we've got particular added salience allied with our huge university expertise in both life sciences and in AI governance. I think if I was to be critical, because I think it's also important to look at where the gaps might be.
In my experiences speaking a lot of summits in different parts of the world. The UK, I think, hasn't been joined up between the Foreign and Commonwealth Office and the foreign, Commonwealth and Development Office and and the former DC Department of Science and Technology. So I think that there hasn't been often you've got decent representatives that that that don't have any experience in multilateral engagement and are therefore quite fearful in engaging, for example, with the Chinese because they're worried about what can I say? What can't I is this a security issue? And so on.
So we need to bring the Foreign Office and Cabinet Office expertise and comfort in dealing with very sensitive issues to the multilateral offer. And I've seen really encouraging signs with the new best replaced de cit with a Cabinet Office now overseeing the AI Security Institute allows a government approach. So I think the omens are are very good that we can have that joined up approach. But I think it's it's absolutely essential that the Cabinet Office actually steers this and ensures that we have a a credible offer, not just on the AI safety side, but also on the AI development side, not just because China is really leading on that, but because developing countries and countries in Central Asia and other parts of the world, they don't see the world through the AI safety lens.
They think, well, if it's mainly the problems caused by companies from US and China, what what's what's our role in that? So you have to engage people from where they are and from much of the world. That's around issues around AI sovereignty and what how they can manage dependencies across the full stack of AI. So I think Lord Bruce has got a question that in many ways follows on from that. Well, I
So we've got real experience of how you actually. Get buy in from around the world for crucial issues. We've got the huge expertise in AI safety from the AI Security Institute and , and so on. And biosafety, as I mentioned before, I think is an area where we've got particular added salience allied with our huge university expertise in both life sciences and in AI governance. I think if I was to be critical, because I think it's also important to look at where the gaps might be.
In my experiences speaking a lot of summits in different parts of the world. The UK, I think, hasn't been joined up between the Foreign and Commonwealth Office and the foreign, Commonwealth and Development Office and and the former DC Department of Science and Technology. So I think that there hasn't been often you've got decent representatives that that that don't have any experience in multilateral engagement and are therefore quite fearful in engaging, for example, with the Chinese because they're worried about what can I say? What can't I is this a security issue? And so on.
So we need to bring the Foreign Office and Cabinet Office expertise and comfort in dealing with very sensitive issues to the multilateral offer. And I've seen really encouraging signs with the new best replaced de cit with a Cabinet Office now overseeing the AI Security Institute allows a government approach. So I think the omens are are very good that we can have that joined up approach. But I think it's it's absolutely essential that the Cabinet Office actually steers this and ensures that we have a a credible offer, not just on the AI safety side, but also on the AI development side, not just because China is really leading on that, but because developing countries and countries in Central Asia and other parts of the world, they don't see the world through the AI safety lens.
They think, well, if it's mainly the problems caused by companies from US and China, what what's what's our role in that? So you have to engage people from where they are and from much of the world. That's around issues around AI sovereignty and what how they can manage dependencies across the full stack of AI. So I think Lord Bruce has got a question that in many ways follows on from that. Well, I
11:19
Lord Bruce of Bennachie
-
-
Lord Bruce of Bennachie
was going to say you've answered a lot of those, but a couple of things . In the July GDP report, unexpectedly, GDP rose by 0.4%, and this was attributed to productivity improvements delivered by AI. That may be just a one off, but if it's actually the first green shoot, it's potentially a very strong one if that was to continue. So I want to link two things. One is the domestic importance of AI and how government prioritise it, linked to our international engagement.
And you've you've sort of indicated you think you have capacity, but for example, we've dismantled our aid budget and the personnel deployment that goes with it. We're reducing our diplomatic service in numbers. So it's a bit of a bee in my bonnet as to whether or not we need to re deploy our diplomatic capacity, particularly the overseas part of it, in terms of looking at emerging economies and what we with our AI capacity can do that's beneficial to them and us, and looking at our like minded allies and dare, where we can jointly both develop safety systems but also develop mutually beneficial technologies.
So how do we use our diplomatic capacity to achieve that? I mean, given that it's changing, how much does it need to change?
And you've you've sort of indicated you think you have capacity, but for example, we've dismantled our aid budget and the personnel deployment that goes with it. We're reducing our diplomatic service in numbers. So it's a bit of a bee in my bonnet as to whether or not we need to re deploy our diplomatic capacity, particularly the overseas part of it, in terms of looking at emerging economies and what we with our AI capacity can do that's beneficial to them and us, and looking at our like minded allies and dare, where we can jointly both develop safety systems but also develop mutually beneficial technologies.
So how do we use our diplomatic capacity to achieve that? I mean, given that it's changing, how much does it need to change?
11:21
11:21
Lord Robertson of Port Ellen
-
-
Lord Robertson of Port Ellen
I'll try it. Really, really great question. I suppose it's part of that whole what's our offer on, on on development. I mean, for a long time , Oda has been overtaken by foreign direct investment and remittances in terms of the volume of its assistance. So it's not really about aid anymore. But we have a tremendous kind of in-kind contribution that we can make. So, I mean, one avenue is our AI Security Institute partnering with the the only AI Security and Safety Institute in Africa at the moment is in Kenya, for example, but partnering with Kenya, partnering with perhaps with Singapore and doing that and with other other African countries spreading out across the African Union.
So there's there's very specific ways in which we can work with, with, with institutions. I think that there it's probably best done in partnership with the different UN agencies that have national capability programs, the UN Development Programme, the, the, the, the different Unesco, the ITU and Unido are the four UN agencies that are all are engaging at a national level in developing countries. So I think that though that is a kind of a key, key way to do it.
And I think that that you're right, that AI will, I hope, continue to drive productivity. In the most optimistic example of that, it will completely change the nature of the debate between the developed and developing countries, things that are holding back agreement on climate change and other things by by bringing in a huge number of resources into developing countries through efficiencies, but developing country priorities, health, education, delivering of digital public services, smart cities and so on.
All of these can see an absolute sea change. And they don't need frontier AI. They can use open source, free open source models. They don't need the latest technology in order to achieve a lot of these economic efficiencies. So we need to prepare for a world. And this, I suppose, goes to the heart of, of, of the Foreign Office restructuring, where we may actually see a huge beneficial uplift in the resources available to all countries through the adoption of, of AI and other technologies.
So we've talked a lot about gloom, but I think there is a possible very positive economic picture. I have done a lot with with Central Asia and Kazakhstan returning there in a few weeks time. And they've got an AI minister, a international advisory panel for the for the president on this area. And they are they're revolutionizing their whole country in ways that the UK almost can't imagine. They are really focusing on that.
Before the Iran war, the UAE and Saudi Arabia were seeing that their path to moving beyond oil and gas was through AI. So I think we are below the surface. It's not necessarily visible on these conversations around the risks. And China, US, there's a whole lot happening on AI at the at the less frontier level, which is actually transforming the world. Click by click, as it were.
So there's there's very specific ways in which we can work with, with, with institutions. I think that there it's probably best done in partnership with the different UN agencies that have national capability programs, the UN Development Programme, the, the, the, the different Unesco, the ITU and Unido are the four UN agencies that are all are engaging at a national level in developing countries. So I think that though that is a kind of a key, key way to do it.
And I think that that you're right, that AI will, I hope, continue to drive productivity. In the most optimistic example of that, it will completely change the nature of the debate between the developed and developing countries, things that are holding back agreement on climate change and other things by by bringing in a huge number of resources into developing countries through efficiencies, but developing country priorities, health, education, delivering of digital public services, smart cities and so on.
All of these can see an absolute sea change. And they don't need frontier AI. They can use open source, free open source models. They don't need the latest technology in order to achieve a lot of these economic efficiencies. So we need to prepare for a world. And this, I suppose, goes to the heart of, of, of the Foreign Office restructuring, where we may actually see a huge beneficial uplift in the resources available to all countries through the adoption of, of AI and other technologies.
So we've talked a lot about gloom, but I think there is a possible very positive economic picture. I have done a lot with with Central Asia and Kazakhstan returning there in a few weeks time. And they've got an AI minister, a international advisory panel for the for the president on this area. And they are they're revolutionizing their whole country in ways that the UK almost can't imagine. They are really focusing on that.
Before the Iran war, the UAE and Saudi Arabia were seeing that their path to moving beyond oil and gas was through AI. So I think we are below the surface. It's not necessarily visible on these conversations around the risks. And China, US, there's a whole lot happening on AI at the at the less frontier level, which is actually transforming the world. Click by click, as it were.
11:24
Speaker 1
-
Speaker 1
Perhaps I could jump in to zoom into two tools in the UK's diplomatic toolbox and AI, drawing on Mr. Dawes fantastic intervention. The first relates to a theme I mentioned earlier in this session, which is the unique role the UK can play in building capacity. Capacity building has come up time and time again in global governance. On AI debates, it is clearly a priority for countries in the global majority.
It is front and centre of China's world AI cooperation organization, but the UK can play a leading international role in building capacity in many of the ways Mr. Dawes detailed. A specific way it can do so speaks to an earlier question raised around preparing for a multitude of different scenarios in the AI governance space. We've seen the emergence of really exciting and innovative future AI scenarios that bring together key policy makers from across sectors and across borders to give them the tools, the awareness, and the networks they need to deal with .
Anything from an AI driven disruption in shipping and trade to an AI national security, an AI driven national security crisis. Now, the UK can work with partners, with other middle powers, with countries in the global majority, to develop these sorts of scenarios. This training to not only improve preparedness, but also ensure that countries facing disruption have the sort of regional and international diplomacy they need. That familiarity, those networks, as we've discussed before, in order to navigate it on the one hand and build governance in conditions of crisis and turbulence on the other.
So a specific tool the UK can use in this emerging diplomatic toolkit on AI looks at scenario based capacity building, and I'd be happy to refer the committee to future work and emerging best practice on this point. The second I would draw your attention to is standards. Now, standards are a relatively dry area of global AI governance, but they are incredibly important. They pertain to the best practice and the standardisation of how, for example, models are tested and evaluated.
How, for example, the security of open source models and the complex supply chains that feed into them is guaranteed. Now, the UK can play a leading international role and broker agreements surrounding, for example, you know, a democratic approach to the development, deployment of AI standards. And I would absolutely push the committee to center this area. As you know, a relatively, I think, important tool in this emerging toolkit that we've discussed just now.
It is front and centre of China's world AI cooperation organization, but the UK can play a leading international role in building capacity in many of the ways Mr. Dawes detailed. A specific way it can do so speaks to an earlier question raised around preparing for a multitude of different scenarios in the AI governance space. We've seen the emergence of really exciting and innovative future AI scenarios that bring together key policy makers from across sectors and across borders to give them the tools, the awareness, and the networks they need to deal with .
Anything from an AI driven disruption in shipping and trade to an AI national security, an AI driven national security crisis. Now, the UK can work with partners, with other middle powers, with countries in the global majority, to develop these sorts of scenarios. This training to not only improve preparedness, but also ensure that countries facing disruption have the sort of regional and international diplomacy they need. That familiarity, those networks, as we've discussed before, in order to navigate it on the one hand and build governance in conditions of crisis and turbulence on the other.
So a specific tool the UK can use in this emerging diplomatic toolkit on AI looks at scenario based capacity building, and I'd be happy to refer the committee to future work and emerging best practice on this point. The second I would draw your attention to is standards. Now, standards are a relatively dry area of global AI governance, but they are incredibly important. They pertain to the best practice and the standardisation of how, for example, models are tested and evaluated.
How, for example, the security of open source models and the complex supply chains that feed into them is guaranteed. Now, the UK can play a leading international role and broker agreements surrounding, for example, you know, a democratic approach to the development, deployment of AI standards. And I would absolutely push the committee to center this area. As you know, a relatively, I think, important tool in this emerging toolkit that we've discussed just now.
11:27
11:27
11:28
11:28
11:28
Lord Lamont of Lerwick
-
-
Lord Lamont of Lerwick
Lord Lamont to the question. Yes. I mean, you talked about the. patchwork pattern that there is. Are we going to see more. convergence and agreement among select. groupings as opposed to large scale agreement via the UN? And we had a given to us the Brookings Institute paper, which actually talked about the need for a. decentralized network approach . Rather
11:28
11:28
11:28
11:28
11:29
11:29
Speaker 1
-
Speaker 1
Shared values. As Mr. Dawes covered, the guiding theme of. sovereignty is front and center of many of these mini lateral agreements and declarations. Now, on the one hand. fragmentation in any area of governance is expected and really should be welcomed. There is no silver bullet in global AI governance. We shouldn't all be governing AI in exactly the same way. And there are important efforts, for example, led by the UN to improve interoperability.
So this weaving together between different, say, regional. approaches or issue based approaches to global AI governance. However, there are areas where. fragmentation is harmful. We've seen over the past few weeks. declarations coming out of the Shanghai Cooperation. Organization coming out of the world AI Cooperation.
So this weaving together between different, say, regional. approaches or issue based approaches to global AI governance. However, there are areas where. fragmentation is harmful. We've seen over the past few weeks. declarations coming out of the Shanghai Cooperation. Organization coming out of the world AI Cooperation.
11:29
11:29
Speaker 1
-
Speaker 1
US led pact, silica various regional groupings. On the one hand, it may be relatively convenient to see these as competitor groupings and absolutely at loggerheads with each other. But on the other, it's important, I think, to accept this kind of regional multitude of different. approaches and seek to build that connective tissue between them. I would refer the committee. absolutely to a recent report by the UN's recently founded AI governance for Humanity Lab, which mapped out the state of interoperability.
So again, a systematic approach to weaving. together the overlap points or the junctures between these different governance. approaches to ensure that harm is reduced, that there's maximum.
So again, a systematic approach to weaving. together the overlap points or the junctures between these different governance. approaches to ensure that harm is reduced, that there's maximum.
11:30
11:30
11:30
11:30
11:30
11:30
11:30
11:30
11:31
11:31
11:31
11:31
11:31
11:31
11:31
Speaker 2
-
Speaker 2
The British Standards Institute can play a really important role. And I think that's one of the real areas for UK influence is actually through standards. And that sets a benchmark. We also have to engage with cenelec, the EU standards organizations, because there is a risk that. Because what the EU AI act determines becomes law for those.
11:31
11:31
Speaker 2
-
Speaker 2
Are affected by the Brussels effect and effectively adopt cenelec. standards as international standards and we need. Therefore, BSI and others and. individual UK. companies need to be engaged in these European standard setting processes as well as the. The global processes. And I think you make a really good point about about the you've got these mini lateral kind of groupings, which are, which are patchwork and fragmented.
I think there does need to be interoperability. There was a useful report that the committee may want to look at, produced by a new UN. university body in Valencia, Spain, on. interoperability across all sorts of different dimensions, from culture to to safety and security. And that's part of the kind of weaving together these. And I think we need to think about how China is engaged with some of the. Western focused organizations in ways that don't harm Western or UK.
competitiveness or, or security.
I think there does need to be interoperability. There was a useful report that the committee may want to look at, produced by a new UN. university body in Valencia, Spain, on. interoperability across all sorts of different dimensions, from culture to to safety and security. And that's part of the kind of weaving together these. And I think we need to think about how China is engaged with some of the. Western focused organizations in ways that don't harm Western or UK.
competitiveness or, or security.
11:33
11:33
11:33
11:33
Speaker 2
-
Speaker 2
With the Global Partnership on AI, it brought in countries like India and Brazil, but China still sits outside of that. So China is a partner with the OECD in certain areas, but not in AI. So finding a way to. connect the. nascent Chinese. standards being projected through their mini laterals with the Western ones. I think this could be a real opportunity with the UK's G20.
presidency to do that. And secondly, Chinese engagement with Casey. So with the kind of the global network of AI, safety and security. institutes renamed names, now China's. attended kind of alongside some of the meetings of that. network in terms of individual academics kind of rocking up and but they haven't been formally involved. So we need trusted dialogues. between the Chinese network of AI safety and security and development.
institutes and Western AI safety and security institutes. So that can't . Come
presidency to do that. And secondly, Chinese engagement with Casey. So with the kind of the global network of AI, safety and security. institutes renamed names, now China's. attended kind of alongside some of the meetings of that. network in terms of individual academics kind of rocking up and but they haven't been formally involved. So we need trusted dialogues. between the Chinese network of AI safety and security and development.
institutes and Western AI safety and security institutes. So that can't . Come
11:34
11:34
Lord Alderdice
-
-
Lord Alderdice
Thanks for whatever you want to pursue that. Yes. Well, Sam, you very kindly. participated in an event I organised in this very room last Thursday on governance of AI, and in. particular because the involvement of the Secretary General of the IPU on the role of. parliamentarians in developing guardrails for AI. So I wonder if I might ask both of you to address which areas of AI governance and the agenda do you see as most likely to produce some kind of agreement, and which.
areas will we find? The prospect for consensus is very limited. If we're going to focus on things we can deliver on, which are more likely to be . Successful and which are more
areas will we find? The prospect for consensus is very limited. If we're going to focus on things we can deliver on, which are more likely to be . Successful and which are more
11:35
11:35
11:35
11:35
11:35
11:35
Speaker 1
-
Speaker 1
Lacking teeth. This is important nonetheless. And on the other side, we can see hard. regulation binding rules. Now we're seeing. likely spotlights of. success on any range of agreements, on any parts of the spectrum. At the moment. Many of this is due to the highly salient policy window we find ourselves in. But let's take a step back and consider how change.
happens when it comes to AI multilateralism. As we've discussed before, as we've covered in our deep dives on different tools and. mechanisms that the UK is involved in, we see the development of consensus best practice at a small level, maybe just a technical forum or an international network or mini lateral grouping, which becomes. platforms and.
happens when it comes to AI multilateralism. As we've discussed before, as we've covered in our deep dives on different tools and. mechanisms that the UK is involved in, we see the development of consensus best practice at a small level, maybe just a technical forum or an international network or mini lateral grouping, which becomes. platforms and.
11:36
11:36
11:36
11:36
11:36
11:36
11:36
11:36
11:36
11:36
11:37
11:37
11:37
11:37
11:37
11:37
11:37
11:38
11:38
11:38
11:38
Speaker 1
-
Speaker 1
Which can then be. platformed to something. concrete and with teeth. This is a vital process. Let's look now at areas where agreement is unlikely. And we've covered this, so I'm not going to spend too long on it. These are areas. laden with. competition and with national values that may be at loggerheads with each other. We've seen the difficulty of brokering mini lateral or even. international agreement on the use of AI in the military domain.
Despite the emergence of many norms and red lines about. unacceptable use at a smaller level. Earlier I mentioned deepfakes as a subset of risks in the information environment and.
Despite the emergence of many norms and red lines about. unacceptable use at a smaller level. Earlier I mentioned deepfakes as a subset of risks in the information environment and.
11:39
11:39
11:39
11:39
11:39
11:39
11:39
11:39
Speaker 2
-
Speaker 2
interest. Anything to add? Thank you. I think one one key area where progress has been made and it will continue is technical standards that we've we've we've talked about that in. If we if we think back to to Miss. Atkinson's division between malicious use accidents and systemic. Risks, I think on the on the.
11:40
11:40
11:40
11:40
11:40
Speaker 2
-
Speaker 2
Are much harder to negotiate. The UN has been negotiating for ten years on on this issue on sometimes colloquially called killer killer robots. And that's partly because they are actively involved in in in Ukraine, Russia, war and other areas. Iran is very opposed to these things. And in the disarmament.
11:40
11:40
11:40
11:40
Speaker 2
-
Speaker 2
And shared risk management frameworks and so on. That has to be multi disciplinary. And it has to be multi partnership with, with involving the private sector to, to, to be successful. The last one I would say is AI for science. I think there's a huge benefit for. humanity to harness the. discoveries that are going to be. cascading our way in, in advances in pure mathematics, in medicine and in, in material science.
And we, I think we have a duty to. ensure that. geopolitics interferes as much as little as. possible in the scientific.
And we, I think we have a duty to. ensure that. geopolitics interferes as much as little as. possible in the scientific.
11:41
11:41
11:42
11:42
11:42
11:42
11:42
Baroness Fraser of Craigmaddie
-
-
Baroness Fraser of Craigmaddie
Which both of you have touched on. You've said that we might be at a time where there are shared. incentives for the private sector, partly to keep the trust of the public and of governments, because if they don't, the vast economic benefits that they have garnered to date will be at risk. But interestingly, we've seen. recently OpenAI, anthropic, and I think Google DeepMind coming together to try and convene.
11:42
11:43
11:43
Baroness Fraser of Craigmaddie
-
-
Baroness Fraser of Craigmaddie
But their view of a standards body isn't necessarily the view of other organizations. And it's back to Lord Ahmed's comment earlier about everybody agrees that something should be done, but it's proving difficult to agree what that something should be and who's whose version we should take. And. And in the world of.
11:43
11:43
11:43
11:43
11:43
11:43
11:43
11:43
11:43
11:43
11:44
11:44
11:44
11:44
Speaker 1
-
Speaker 1
I can kick off a fantastic question. Thank you so much, Baroness Fraser. We have been grappling with the role of the private sector in global governance for decades. You know, we've had scholars like Susan Strange in the early. 90s talking about the retreat of the state, you know, global. governance or international relations, formally as the playground of sovereign states being. challenged by these incredibly powerful multinational corporations. So many of the anxieties that you note in your question are ones that we've been facing in other areas of governance for a long time.
But over the past few years, we're encountering. unprecedented market power of these new AI companies and older. technology companies. disproportionate influence over capability. The frontier is.
But over the past few years, we're encountering. unprecedented market power of these new AI companies and older. technology companies. disproportionate influence over capability. The frontier is.
11:45
11:45
11:45
11:45
11:45
11:45
11:45
11:45
11:45
11:45
11:45
11:45
Speaker 1
-
Speaker 1
This division absolutely has to break down, because what we're seeing is a great deal of. hybridity in that approaches to, say, documenting what models can do. So model cards are developed by the developers of those models themselves. Best practice is. developed among, say, companies and the technical community. And then eventually you'll see some of this best practice. reflected in regulation. So that is to say, there is a huge. influence over the global governance of AI, noting that if you govern a model, it of course has.
cross-border and. cross-domain impacts. None of this governance is stuck to one policy area, just one model or just one jurisdiction. Now, we've seen private sector. crowding when it comes to many of the new governance. proposals swirling around over the past summer. Dario Ahmed's essay on Pacing the frontier. Many of you will be familiar with calls for an. independent third party evaluator network.
We've seen Demis Hassabis of Google DeepMind calling for a US led AI standards.
cross-border and. cross-domain impacts. None of this governance is stuck to one policy area, just one model or just one jurisdiction. Now, we've seen private sector. crowding when it comes to many of the new governance. proposals swirling around over the past summer. Dario Ahmed's essay on Pacing the frontier. Many of you will be familiar with calls for an. independent third party evaluator network.
We've seen Demis Hassabis of Google DeepMind calling for a US led AI standards.
11:46
11:46
Speaker 1
-
Speaker 1
Sam Altman and Elon Musk throwing their weight. behind Dario Ahmad's proposals as well. On the one hand. On the other hand, we've seen, you know, Nvidia and Jensen Huang and Musk and excuse me, Meta's Mark Zuckerberg coming out in. opposition to these calls for regulation. So I suppose the puzzle that we find ourselves in is what to do here. You know, perhaps the private sector doesn't need a formalized role in global AI governance.
I would argue they are already there. They are already shaping the debate, mediating the discourse, creating many of the rooms that these critical. decisions about the governance of technology are taking place in the puzzle that we face. And I think the critical question that we need to answer and governments need a coherent approach is, is how to ensure. accountability and maximize governance. agency when coming into.
I would argue they are already there. They are already shaping the debate, mediating the discourse, creating many of the rooms that these critical. decisions about the governance of technology are taking place in the puzzle that we face. And I think the critical question that we need to answer and governments need a coherent approach is, is how to ensure. accountability and maximize governance. agency when coming into.
11:47
11:47
Speaker 1
-
Speaker 1
Or perhaps. partnership with these new. concentrations of corporate power at the global level. And there are ways to doing so. My bottom line here would be that the global governance of AI does need to be designed with private sector in the room. There's very little that we can do without them, but it can't be a relationship based simply on voluntary agreements, on. trust and on goodwill.
There has to be a degree of agency and of reciprocity. And we see the. co-development of trusted.
There has to be a degree of agency and of reciprocity. And we see the. co-development of trusted.
11:48
11:48
Speaker 1
-
Speaker 1
Of high trust technical. instruments in different areas of global AI governance. So we know we can do this, but it is a matter of incubating, and it's also a matter for the UK to consider where. coordinating with other middle powers who maybe don't have access to the most powerful frontier. capabilities and are encountering these.
11:48
11:48
Speaker 1
-
Speaker 1
AI companies where they might be able to coordinate collectively on things like access and. potentially on soft regulation. I would love us to return to the point, and the. proposal for setting up an. independent network of third party evaluators. So these are independent. evaluators who would go into AI companies and test the models to see, for example, different safety and security risks that can be. mitigated before they're released.
The UK can play a central role in any of these governance proposals. We have the technical expertise and diplomatic. capital to do so, and this could be a fantastic route to really approach, again, this critical relationship with the private sector and global AI. governance with a two .
The UK can play a central role in any of these governance proposals. We have the technical expertise and diplomatic. capital to do so, and this could be a fantastic route to really approach, again, this critical relationship with the private sector and global AI. governance with a two .
11:49
Baroness Fraser of Craigmaddie
-
-
Baroness Fraser of Craigmaddie
Principles I mentioned earlier. accountability and agency. I just can I just quickly come back to you on that last point? It's all very well to say, let's send people in to look at the safety aspects of a model before it's released. Isn't there also an issue about the companies who are developing it are saying, once we've released it and you're deploying it, it's not my it was safe when it left me, but you're deploying it in a way you've chosen to.
So so . It's
So so . It's
11:50
Speaker 1
-
Speaker 1
not just as simple as as that. Absolutely. Thank you so much for raising that point. Testing and. evaluation is by no means a silver bullet. A lot of the time, the. environment or the part of the. economy or the community where a model is. deployed will bring up a variety of different risks, which are really difficult to test for at the testing and evaluation stage.
This is why it's so important to promote.
This is why it's so important to promote.
11:50
11:50
11:50
Speaker 1
-
Speaker 1
A risk. threshold or notification. threshold for if model behaviour is causing X amount of harm. This is often called technical intervention or a kill switch, which I'd be happy to explain further, or refer the committee to some technical work on. But I would. underline that in order to test and evaluate the safety of models.
11:51
11:51
11:51
11:51
11:51
11:51
Speaker 2
-
Speaker 2
I will take Baroness Fraser, taking your last point first, it really points to the need for clear attribution of. responsibility among different parts of the whole supply chain, as it were. One example is open source models. If they're. modified and harm is caused, whether it's the original open source developer that's responsible or for the person that did, the company that did the modification. And how do you prove that Agentic AI has the same thing? And so actually having a, an agent.
identification agent. tracking is really important. So, you know, where is an agent at any particular time, what it's done. The problem with the kill switch, for example, it's a great idea and maybe necessary, but is that an agent may be embedded in a very sensitive technology at that time. So what you really need, you need maybe a kill switch, but you also need the ability to roll back. permissions over over time. So take it back to when it didn't start acting in a non-aligned or misaligned way.
Answering your original question , the traditional way that Western. companies have engaged in multilateral. processes has largely been voluntary through, for example, the the Frontier Model Forum through the G7. Hiroshima AI. process and through the standards bodies that we've talked about. And it's been really exchanging best practice, and it's all been valuable. They've also made. commitments as a result of the Bletchley summit to the follow up to Bletchley, which was Seoul.
So they made commitments to the Seoul summit. Chinese companies in that case also did. Chinese companies are not part of the Frontier Model Forum. They're not part of the other. processes except for the standards ones. So it was one of the few areas where they could also put forward voluntary commitments.
identification agent. tracking is really important. So, you know, where is an agent at any particular time, what it's done. The problem with the kill switch, for example, it's a great idea and maybe necessary, but is that an agent may be embedded in a very sensitive technology at that time. So what you really need, you need maybe a kill switch, but you also need the ability to roll back. permissions over over time. So take it back to when it didn't start acting in a non-aligned or misaligned way.
Answering your original question , the traditional way that Western. companies have engaged in multilateral. processes has largely been voluntary through, for example, the the Frontier Model Forum through the G7. Hiroshima AI. process and through the standards bodies that we've talked about. And it's been really exchanging best practice, and it's all been valuable. They've also made. commitments as a result of the Bletchley summit to the follow up to Bletchley, which was Seoul.
So they made commitments to the Seoul summit. Chinese companies in that case also did. Chinese companies are not part of the Frontier Model Forum. They're not part of the other. processes except for the standards ones. So it was one of the few areas where they could also put forward voluntary commitments.
11:53
11:53
11:53
11:53
Speaker 2
-
Speaker 2
Was that the Chinese companies who did put put their head above the parapet and made commitments that were in the Seoul. framework were then hit by export controls by the US. And so I've been to China three times in the last nine months, talking to AI experts there in. academia and elsewhere. And they they feel that they they're walking very on very sensitive ground on wanting to accord with what the Chinese government want and need and what the international community want.
So I think it's quite difficult for those private actors. And when we talk about the private sector, given that Chinese capabilities on the frontier are between six months and two years behind. However, you according to different estimates, they are going to be very close to where we are. So they need to be involved. And the Chinese government already placed certain obligations and algorithmic.
So I think it's quite difficult for those private actors. And when we talk about the private sector, given that Chinese capabilities on the frontier are between six months and two years behind. However, you according to different estimates, they are going to be very close to where we are. So they need to be involved. And the Chinese government already placed certain obligations and algorithmic.
11:54
11:54
11:55
11:55
11:55
11:55
11:55
11:55
11:55
11:55
11:55
11:56
11:56
Speaker 1
-
Speaker 1
International Network for advanced AI. Measurements and Science in names. If I'm getting the acronym correctly, which is the renaming of the International Network of AI Safety institutes. Again, this is a critical network for setting the. science around AI safety, developing best practice on testing and .
11:56
11:56
11:56
11:56
11:56
11:56
11:56
11:56
Speaker 1
-
Speaker 1
again, are really valuable architectures and could be. bolstered if we wanted to action some of the governance proposals that we've been discussing today. One of the leading. scientists in AI governance, in technical AI governance, and I'll mention also a Chatham House associate Fellow, Stuart Russell runs the International Association for Safe and Ethical AI, which is a critical technical gathering, setting best. practice on different technical governance mechanisms, and then scaling this internationally.
We've seen different track. 1.5 efforts and scientists to scientist dialogues and AI safety space. And of course the successor to the Bletchley summit. So that's the Geneva AI summit taking place in June next year. I mention all of these different. architectures to say there are concrete. opportunities to better resource them. There are strategic. opportunities to improve. interoperability between them and to set up permanent mechanisms.
When I was in Geneva at the UN Global Dialogue, there were some fascinating. proposals to give the Geneva AI summit a permanent. secretariat that can, for example, track progress and. implementation around some of the voluntary commitments that Mr. Dawes mentioned. For example, on Frontier Technology, Frontier Company AI safety commitments. Now, it's worth learning from what has worked well in the past, and collecting a sort of patchwork of governance analogues from other areas.
We've mentioned the.
We've seen different track. 1.5 efforts and scientists to scientist dialogues and AI safety space. And of course the successor to the Bletchley summit. So that's the Geneva AI summit taking place in June next year. I mention all of these different. architectures to say there are concrete. opportunities to better resource them. There are strategic. opportunities to improve. interoperability between them and to set up permanent mechanisms.
When I was in Geneva at the UN Global Dialogue, there were some fascinating. proposals to give the Geneva AI summit a permanent. secretariat that can, for example, track progress and. implementation around some of the voluntary commitments that Mr. Dawes mentioned. For example, on Frontier Technology, Frontier Company AI safety commitments. Now, it's worth learning from what has worked well in the past, and collecting a sort of patchwork of governance analogues from other areas.
We've mentioned the.
11:58
11:58
Speaker 1
-
Speaker 1
Which many folks borrow from when they think about the potential future role of an. international scientific panel on AI. stating the state of risk and setting out how different technical. specialists are. mitigating them and tracking them, and developing some sort of information. sharing in order to again set that global best practice. I would also draw the committee's attention to mechanisms like the Financial Stability Board, the. scenarios for future planning that the World Health.
Organization does, and the International Atomic Energy Agency. Again, these are not. perfect analogues. They shouldn't be copy and pasted into the complexity and tempo of the global AI governance landscape. But I would reiterate that as we're seeking not to reinvent the wheel and thinking about promising governance. architectures that we can strengthen and better resource, we have a history guidebook of. lessons not only for governance mechanisms that might work, but many governance mechanisms that have been developed in times of great.
crisis and.
Organization does, and the International Atomic Energy Agency. Again, these are not. perfect analogues. They shouldn't be copy and pasted into the complexity and tempo of the global AI governance landscape. But I would reiterate that as we're seeking not to reinvent the wheel and thinking about promising governance. architectures that we can strengthen and better resource, we have a history guidebook of. lessons not only for governance mechanisms that might work, but many governance mechanisms that have been developed in times of great.
crisis and.
11:59
11:59
11:59
11:59
11:59
12:00
12:00
Speaker 2
-
Speaker 2
To house because it's so far reaching and general purpose. It reaches in a way that. interconnects issues in ways that humans eventually won't even be able to understand, because AI will be able to work out the the interconnections better than us as it gets more intelligent. But it's human. nature and or political nature that we tend to be quite siloed in our approach to things. So governments have a Ministry of Health and Ministry of Defence and so on.
And so that's. reflected in the international. architecture at the moment with the different UN bodies, NATO and so on. So I think the most likely direction of travel, one where there's the most.
And so that's. reflected in the international. architecture at the moment with the different UN bodies, NATO and so on. So I think the most likely direction of travel, one where there's the most.
12:00
12:00
12:00
12:01
12:01
12:01
Speaker 2
-
Speaker 2
The International Labour. Organization will take the lead for the health aspects of AI. The World Health. Organization will, will, will take the lead, and so on. The challenge at the moment is that the multilateral system is coming under the. The international organizations are coming under a huge pressure to deliver more for less in the future, because donors money is also limited. And so the next secretary general, who will be appointed somewhere sometime between today and.
31st of December of this year, she or he will have to. decide how how to embrace AI as a.
31st of December of this year, she or he will have to. decide how how to embrace AI as a.
12:01
12:01
Speaker 2
-
Speaker 2
And whether, as with Guterres, it was something that was quite central and what kind of architecture they want to recommend, how much political capital they want to put into, into setting up new systems for the bits of AI that don't fall neatly into a health basket or a security basket, or so on. There's been. various talk or attempts, as Mrs. Wilkinson. Wilkinson said to. To develop, for example, a.
IAEA for AI, which would have enforcement powers to require states to do something that was considered by the UN Advisory Board on AI, and they decided not to put it forward. There's no appetite at the moment for a UN body with enforcement powers to tell countries or companies what to do, for understandable reasons. So we got a bit of a challenge. I think that we may end up with a body like that in a few years time, but I think the the.
precondition for that would be a major.
IAEA for AI, which would have enforcement powers to require states to do something that was considered by the UN Advisory Board on AI, and they decided not to put it forward. There's no appetite at the moment for a UN body with enforcement powers to tell countries or companies what to do, for understandable reasons. So we got a bit of a challenge. I think that we may end up with a body like that in a few years time, but I think the the.
precondition for that would be a major.
12:03
12:03
Speaker 2
-
Speaker 2
major cyber attack that was so. significantly economically. disruptive that there was a willingness to. relinquish a bit of sovereignty in order to set up such a body. But we're not there yet. And heaven forbid, you know, that the actual crisis is not a lot worse than than a simple economic disruption. But it points to the to the real. difficulties in designing the architecture.
I mean, my one guidance would be that that for the committee in considering it is don't start from an. organogram of what looks sensible on paper. It's actually about. multilateral reform is around aligning very different.
I mean, my one guidance would be that that for the committee in considering it is don't start from an. organogram of what looks sensible on paper. It's actually about. multilateral reform is around aligning very different.
12:03
12:03
12:03
12:03
12:03
12:03
12:03
12:03
12:03
12:03
12:04
12:04
Speaker 2
-
Speaker 2
It prevents the proliferation of nuclear weapons in return. It supports the peaceful development of nuclear technology. I could see a new agency in the future. monitoring the risks from AI in real time, but at the same time supporting developing countries with their AI. development in areas like cyber security. I think we are for geopolitical reasons, the priority should be for the UK to work with our close security. partners like NATO, like G7 countries and so on.
There's some things where I think the global framing, it can be.
There's some things where I think the global framing, it can be.
12:04
12:04
12:04
Baroness Blackstone
-
-
Baroness Blackstone
So there's different different configurations. Thank you very much. Just very quick follow up. I mean, we've said don't reinvent the wheel, although you have just sort of come up with one possible longer term. solution to getting better global governance. I'm a bit sort of stuck between don't invent the wheel, keep going as we are. And yet there's a gridlock in global governance and you keep using that phrase.
So somehow or other, we have to find our . Way
So somehow or other, we have to find our . Way
12:05
12:05
Speaker 2
-
Speaker 2
Through those two positions. I think we need to be ambitious and say, look, this would be the ideal and see if we can push it and if the UK can can help garner some agreement. And I think it will be. incremental in nature. It'll be built up through, first of all, private sector voluntary agreements and. regulations already having some effect. The the state laws in California that have been passed some in recent weeks, and the Brussels effect of the EU AI act, and in particular now the general purpose AI provisions which have come into force while the other.
EU AI act provisions have been postponed because of the omnibus AI initiative of the EU. So you're seeing. companies beginning to bring in. safety controls, or at least safety declarations and. transparency that meet both. California's requirements and the EU's requirements. So I think we're going to see not a perfect evolution, but you're going to see different regulatory steps. And it was quite it was quite noticeable that the the vice president of the the US did did strike a.
EU AI act provisions have been postponed because of the omnibus AI initiative of the EU. So you're seeing. companies beginning to bring in. safety controls, or at least safety declarations and. transparency that meet both. California's requirements and the EU's requirements. So I think we're going to see not a perfect evolution, but you're going to see different regulatory steps. And it was quite it was quite noticeable that the the vice president of the the US did did strike a.
12:06
12:07
12:07
12:07
12:07
12:07
12:07
12:07
Speaker 2
-
Speaker 2
allies to you, given the time he spent in China? Okay. I mean, the Chinese approach is it's a focus on sovereignty. It's a focus on control. It's a focus on. development . And the world AI Cooperation. Organization seeks to be an international version of what's happening at home. So if you look in China, there's a fifth five year plan and their own AI plus initiative, which is about.
implementing AI into. every economic and social sphere in China. So they're offering to the. world is a low cost, open source, full stack.
implementing AI into. every economic and social sphere in China. So they're offering to the. world is a low cost, open source, full stack.
12:08
12:08
12:08
12:08
12:08
12:08
12:09
12:09
12:09
12:09
12:09
12:09
Speaker 2
-
Speaker 2
Democratic values and so on. I think the UK could even consider joining the world AI Cooperation Organization, depending on what kind of formulation and then influencing it from the inside. So I think we need to be nimble in our approach to these things and, and, and not fear everything China does as being. purely a China strategic play. I think partly it will be and partly it will be a genuine desire to improve the lot of the developing world.
And I think when we. When we when we meet with Chinese.
And I think when we. When we when we meet with Chinese.
12:09
12:09
12:10
12:10
12:10
12:10
Speaker 1
-
Speaker 1
Thank you. You want to add briefly, the UK can play a really important role in supporting interoperability. Again. between anything that is agreed and platformed using China's Waco and other. regional or even multilateral initiatives. The place to do so is. absolutely the United Nations. You've seen them take bold. strides pitching the UN global dialogue as a sort of dialogue of dialogues, a hub of different.
partnerships and initiatives. It is in the UK strategic interest to support. interoperability on the standards front is a valuable way to doing so, but also in diplomatic.
partnerships and initiatives. It is in the UK strategic interest to support. interoperability on the standards front is a valuable way to doing so, but also in diplomatic.
12:11
12:11
12:11
12:11
12:11
12:11
12:11
12:11
12:11
12:11
12:11
12:11
12:12
12:12
12:12
Speaker 2
-
Speaker 2
Mr. Dawes has outlined. Thank you. And my answer would be very similar. I'm really glad that we're very aligned. The UK has a huge offer, a huge contribution we can make in AI safety, AI. security and AI for science and diplomatically. I think really perfecting our our different engagement strategies. I think we need a warm, warm relationship with the US with the aim of.
ensuring that US frontier models are accessible to the UK AI Security Institute. We need. regulatory alignment, not not not identical, but alignment with the EU to the extent that it will allow UK. businesses to to prosper and to further our objectives and differ where that is of advantage. And we need engagement with China on a whole series of tracks, particularly between our AI Safety Institute and their Security Institute and their equivalent, but also, I think, as a possible bridge between China and the West.
More broadly, I think we. occupy a uniquely beneficial position outside of the EU with good connections to the the US and with. excellent connections across the Commonwealth. The last area for diplomatic engagement is what is our offer to the rest of the world. And there I think we need tailored plans for Central Asia, Southeast Asia, Latin America, the Gulf and . Africa. And there will be slightly different contributions we can make.
But I think these are not things that will cost a lot of money. I think there's tremendous amount of. expertise in the wider UK. ecosystem that we can contribute to, ensuring that AI helps humanity to flourish, as well as addresses the risks . And
ensuring that US frontier models are accessible to the UK AI Security Institute. We need. regulatory alignment, not not not identical, but alignment with the EU to the extent that it will allow UK. businesses to to prosper and to further our objectives and differ where that is of advantage. And we need engagement with China on a whole series of tracks, particularly between our AI Safety Institute and their Security Institute and their equivalent, but also, I think, as a possible bridge between China and the West.
More broadly, I think we. occupy a uniquely beneficial position outside of the EU with good connections to the the US and with. excellent connections across the Commonwealth. The last area for diplomatic engagement is what is our offer to the rest of the world. And there I think we need tailored plans for Central Asia, Southeast Asia, Latin America, the Gulf and . Africa. And there will be slightly different contributions we can make.
But I think these are not things that will cost a lot of money. I think there's tremendous amount of. expertise in the wider UK. ecosystem that we can contribute to, ensuring that AI helps humanity to flourish, as well as addresses the risks . And
12:14
12:14