Data (Use and Access) Bill [HL]

Lord Kamall Excerpts

Committee stage
Monday 16th December 2024 Grand Committee

[Lord Thomas of Cwmgiedd (CB)]

My Lords, I rise briefly, first, to thank everyone who has spoken so eloquently about the importance of automated decision-making, in particular its importance to public trust and the importance of human intervention. The retrograde step of watering down Article 22 is to be deplored. I am therefore grateful to the noble Lord, Lord Clement-Jones, for putting forward that this part of the Bill should not stand part. Secondly, the specific amendment that I have laid seeks to retain the broader application of human intervention for automated decision-making where it is important. I can see no justification for that watering down, particularly when there is such uncertainty about the scope that AI may bring to what can be done by automated decision-making.

[Lord Kamall (Con)]

My Lords, in speaking to this group of amendments I must apologise to the Committee that, when I spoke last week, I forgot to mention my interests in the register, specifically as an unpaid adviser to the Startup Coalition. For Committee, noble Lords will realise that I have confined myself to amendments that may be relevant to our healthcare and improving that.

I will speak to Amendments 111 and 116 in the names of my noble friends Lord Camrose and Lord Markham, and Amendment 115 from my noble friend Lord Lucas and the noble Lords, Lord Clement-Jones and Lord Knight of Weymouth, as well as other amendments, including from my noble friend Lord Holmes—I will probably touch on most amendments in this group. To illustrate my concerns, I return to two personal experiences that I shared during debate on the Data Protection and Digital Information Bill. I apologise to noble Lords who have heard these examples previously, but they illustrate the points being made in discussing this group of amendments.

A few years ago, when I was supposed to be travelling to Strasbourg, my train to the airport got delayed. My staff picked me up, booked me a new flight and drove me to the airport. I got to the airport with my new boarding pass and scanned it to get into the gate area, but as I was about to get on the flight, I scanned my pass again and was not allowed on the flight. No one there could explain why, having been allowed through security, I was not allowed on the flight. To cut a long story short, after two hours of being gaslighted by four or five staff, with them not even saying that they could not explain things to me, I eventually had to return to the check-in desk—this was supposed to be avoided by all the automation—to ask what had happened. The airline claimed that it had sent me an email that day. The next day, it admitted that it had not sent me an email. It then explained what had happened by saying that a flag had gone off in its system. That was simply the explanation.

This illustrates the point about human intervention, but it is also about telling customers and others what happens when something goes wrong. The company clearly had not trained its staff in how to speak to customers or in transparency. Companies such as that airline get away with this sort of disgraceful behaviour all the time, but imagine if such technology were being used in the NHS. Imagine the same scenario: you turn up for an operation, and you scan your barcode to enter the hospital—possibly even the operating theatre—but you are denied access. There must be accountability, transparency and human intervention, and, in these instances, there has to be human intervention immediately. These things are critical.

I know that this Bill makes some sort of differentiation between more critical and less critical ADM, but let me illustrate my point with another example. A few years ago, I paid for an account with one of those whizzy fintech banks. Its slogan was: “We are here to make money work for everyone”. I downloaded the app and filled out the fields, then a message popped up telling me, “We will get back to you within 48 hours”. Two weeks later, I got a message on the app saying that I had been rejected and that, by law, the bank did not have to explain why. Once again, I ask noble Lords to imagine. Imagine Monzo’s technology being used on the NHS app, which many people currently use for repeat prescriptions or booking appointments. What would happen if you tried to book an appointment but you received a message saying, “Your appointment has been denied and, by law, we do not have to explain why”? I hope that we would have enough common sense to ensure that there is human intervention immediately.

I realise that the noble Lord, Lord Clement-Jones, has a Private Member’s Bill on this issue—I am sorry that I have not been able to take part in those debates—but, for this Bill, I hope that the two examples I have just shared illustrate the point that I know many noble Lords are trying to make in our debate on this group of amendments. I look forward to the response from the Minister.

[Viscount Camrose (Con)]

I thank all noble Lords who have spoken. I must confess that, of all the groups we are looking at today, I have been particularly looking forward to this one. I find this area absolutely fascinating.

Let me begin in that spirit by addressing an amendment in my name and that of my noble friend Lord Markham and I ask the Government and all noble Lords to give it considerable attention. Amendment 111 seeks to insert the five principles set out in the AI White Paper published by the previous Government and to require all those participating in ADM—indeed, all forms of AI—to have due regard for them. They are:

“safety, security and robustness, appropriate transparency and explainability, fairness, accountability and governance, and contestability and redress”.

These principles for safe AI are based on those originally developed with the OECD and have been the subject of extensive consultation. They have been refined and very positively received by developers, public sector organisations, private sector organisations and civil society. They offer real safeguards against the risks of AI while continuing to foster innovation.

I will briefly make three brief points to commend their inclusion in the Bill, as I have described. First, the Bill team has argued throughout that these principles are already addressed by the principles of data protection and so are covered in the Bill. There is overlap, of course, but I do not agree that they are equivalent. Data protection is a significant concern in AI but the risks and, indeed, the possibilities of AI go far further than data protection. We simply cannot entrust all our AI risks to data protection principles.

Secondly, I think the Government will point to their coming AI Bill and suggest that we should wait for that before we move significantly on AI. However, in practice all we have to go on about the Bill—I recognise that Ministers cannot describe much of it now—is that it will focus on the largest AI labs and the largest models. I assume it will place existing voluntary agreements on a statutory footing. In other words, we do not know when the Bill is coming, but this approach will allow a great many smaller AI fish to slip through the net. If we want to enshrine principles into law that cover all use of AI here, this may not quite be the only game in town, but it is certainly the only all-encompassing, holistic game in town likely to be positively impactful. I look forward to the Minister’s comments on this point.