Online Safety Bill (Twelfth sitting)

Kirsty Blackman Excerpts

Committee stage
Thursday 16th June 2022 Public Bill Committees

[Alex Davies-Jones (Pontypridd) (Lab)]

Under this chapter, Ofcom will have the power to direct companies to use accredited technology to identify child sexual exploitation and abuse content, whether communicated publicly or privately by means of a service, and to remove that content quickly. Colleagues will be aware that the Internet Watch Foundation is one group that assists companies in doing that by providing them with “hashes” of previously identified child sexual abuse material in order to prevent the upload of such material to their platforms. That helps stop the images of victims being recirculated again and again. Tech companies can then notify law enforcement of the details of who has uploaded the content, and an investigation can be conducted and offenders sharing the content held to account.

Those technologies are extremely accurate and, thanks to the quality of our datasets, ensure that companies are detecting only imagery that is illegal. There are a number of types of technology that Ofcom could consider accrediting, including image hashing. A hash is a unique string of letters and numbers that can be applied to an image and matched every time a user attempts to upload a known illegal image to a platform.

PhotoDNA is another type, created in 2009 in a collaboration between Microsoft and Professor Hany Farid at the University of Berkeley. PhotoDNA is a vital tool in the detection of CSEA online. It enables law enforcement, charities, non-governmental organisations and the internet industry to find copies of an image even when it has been digitally altered. It is one of the most important technical developments in online child protection. It is extremely accurate, with a failure rate of one in 50 billion to 100 billion. That gives companies a high degree of certainty that what they are removing is illegal, and a firm basis for law enforcement to pursue offenders.

Lastly, there is webpage blocking. Most of the imagery that the Internet Watch Foundation removes from the internet is hosted outside the UK. While it is waiting for removal, it can disable public access to an image or webpage by adding it to our webpage blocking list. That can be utilised by search providers to de-index known webpages containing CSAM. I therefore ask the Minister, as we continue to explore this chapter, to confirm exactly how such technologies can be utilised once the Bill receives Royal Assent.

Labour welcomes clause 105, which confirms, in subsection (2), that where a service provider is already using technology on a voluntary basis but it is ineffective, Ofcom can still intervene and require a service provider to use a more effective technology, or the same technology in a more effective way. It is vital that Ofcom is given the power and opportunity to intervene in the strongest possible sense to ensure that safety online is kept at the forefront.

However, we do require some clarification, particularly on subsections (9) and (10), which explain that Ofcom will only be able to require the use of tools that meet the minimum standards for accuracy for detecting terrorism and/or CSEA content, as set out by the Secretary of State. Although minimum standards are of course a good thing, can the Minister clarify the exact role that the Secretary of State will have in imposing these minimum standards? How will this work in practice?

Once again, Labour does not oppose clause 106 and we have not sought to amend it at this stage. It is vital that Ofcom has the power to revoke a notice under clause 103(1) if there are reasonable grounds to believe that the provider is not complying with it. Only with these powers can we be assured that service providers will be implored to take their responsibilities and statutory duties, as outlined in the Bill, seriously.

[Kirsty Blackman (Aberdeen North) (SNP)]

I have a few questions, concerns and suggestions relating to these clauses. I think it was the hon. Member for Don Valley who asked me last week about the reports to the National Crime Agency and how that would work—about how, if a human was not checking those things, there would be an assurance that proper reports were being made, and that scanning was not happening and reports were not being made when images were totally legal and there was no problem with them. [Interruption.] I thought it was the hon. Member for Don Valley, although it may not have been. Apologies—it was a Conservative Member. I am sorry for misnaming the hon. Member.

The hon. Member for Pontypridd made a point about the high level of accuracy of the technologies. That should give everybody a level of reassurance that the reports that are and should be made to the National Crime Agency on child sexual abuse images will be made on a highly accurate basis, rather than a potentially inaccurate one. Actually, some computer technology—particularly for scanning for images, rather than text—is more accurate than human beings. I am pleased to hear those particular statistics.

Queries have been raised on this matter by external organisations—I am particularly thinking about the NSPCC, which we spoke about earlier. The Minister has thankfully given a number of significant reassurances about the ability to proactively scan. External organisations such as the NSPCC are still concerned that there is not enough on the face of the Bill about proactive scanning and ensuring that the current level of proactive scanning is able—or required—to be replicated when the Bill comes into action.

During an exchange in an earlier Committee sitting, the Minister gave a commitment—I am afraid I do not have the quote—to being open to looking at amending clause 103. I am slightly disappointed that there are no Government amendments, but I understand that there has been only a fairly short period; I am far less disappointed than I was previously, when the Minister had much more time to consider the actions he might have been willing to take.

The suggestion I received from the NSPCC is about the gap in the Bill regarding the ability of Ofcom to take action. These clauses allow Ofcom to take action against individual providers about which it has concerns; those providers will have to undertake duties set out by Ofcom. The NSPCC suggests that there could be a risk register, or that a notice could be served on a number of companies at one time, rather than Ofcom simply having to pick one company, or to repeatedly pick single companies and serve notices on them. Clause 83 outlines a register of risk profiles that must be created by Ofcom. It could therefore serve notice on all the companies that fall within a certain risk profile or all the providers that have common functionalities.

If there were a new, emerging concern, that would make sense. Rather than Ofcom having to go through the individual process with all the individual providers when it knows that there is common functionality—because of the risk assessments that have been done and Ofcom’s oversight of the different providers—it could serve notice on all of them in one go. It could not then accidentally miss one out and allow people to move to a different platform that had not been mentioned. I appreciate the conversation we had around this issue earlier, and the opportunity to provide context in relation to the NSPCC’s suggestions, but it would be great if the Minister would be willing to consider them.

I have another question, to which I think the Minister will be able to reply in the affirmative, which is on the uses of the technology as it evolves. We spoke about that in an earlier meeting. The technology that we have may not be what we use in the future to scan for terrorist-related activity or child sexual abuse material. It is important that the Bill adequately covers future conditions. I think that it does, but will the Minister confirm that, as technology advances and changes, these clauses will adequately capture the scanning technologies that are required, and any updates in the way in which platforms work and we interact with each other on the internet?

I have fewer concerns about future-proofing with regard to these provisions, because I genuinely think they cover future conditions, but it would be incredibly helpful and provide me with a bit of reassurance if the Minister could confirm that. I very much look forward to hearing his comments on clause 103.

[The Parliamentary Under-Secretary of State for Digital, Culture, Media and Sport (Chris Philp)]

Let me start by addressing some questions raised by hon. Members, beginning with the last point made by the hon. Member for Aberdeen North. She sought reconfirmation that the Bill will keep up with future developments in accredited technology that are not currently contemplated. The answer to her question can be found in clause 105(9), in which the definition of accredited technology is clearly set out, as technology that is

“accredited (by OFCOM or another person appointed by OFCOM) as meeting minimum standards of accuracy”.

That is not a one-off determination; it is a determination, or an accreditation, that can happen from time to time, periodically or at any point in the future. As and when new technologies emerge that meet the minimum standards of accuracy, they can be accredited, and the power in clause 103 can be used to compel platforms to use those technologies. I hope that provides the reassurance that the hon. Member was quite rightly asking for.

The shadow Minister, the hon. Member for Pontypridd, asked a related question about the process for publishing those minimum standards. The process is set out in clause 105(10), which says that Ofcom will give advice to the Secretary of State on the appropriate minimum standards, and the minimum standards will then be

“approved…by the Secretary of State, following advice from OFCOM.”

We are currently working with Ofcom to finalise the process for setting those standards, which of course will need to take a wide range of factors into account.

Let me turn to the substantive clauses. Clause 103 is extremely important, because as we heard in the evidence sessions and as Members of the Committee have said, scanning messages using technology such as hash matching, to which the shadow Minister referred, is an extremely powerful way of detecting CSEA content and providing information for law enforcement agencies to arrest suspected paedophiles. I think it was in the European Union that Meta—particularly Facebook and Facebook Messenger—stopped using this scanner for a short period time due to misplaced concerns about privacy laws, and the number of referrals of CSEA images and the number of potential paedophiles who were referred to law enforcement dropped dramatically.

A point that the hon. Member for Aberdeen North and I have discussed previously is that it would be completely unacceptable if a situation arose whereby these messages—I am thinking particularly about Facebook Messenger—did not get scanned for CSEA content in a way that they do get scanned today. When it comes to preventing child sexual exploitation and abuse, in my view there is no scope for compromise or ambiguity. That scanning is happening at the moment; it is protecting children on a very large scale and detecting paedophiles on quite a large scale. In my view, under no circumstances should that scanning be allowed to stop. That is the motivation behind clause 103, which provides Ofcom with the power to make directions to require the use of accredited technology.

As the hon. Member for Aberdeen North signalled in her remarks, given the importance of this issue the Government are of course open to thinking about ways in which the Bill can be strengthened if necessary, because we do not want to leave any loopholes. I urge any social media firms watching our proceedings never to take any steps that degrade or reduce the ability to scan for CSEA content. I thank the hon. Member for sending through the note from the NSPCC, which I have received and will look at internally.

--- Later in debate ---

[Alex Davies-Jones]

We welcome clause 104, but have tabled some important amendments that the Minister should closely consider. More broadly, the move away from requiring child sexual exploitation and abuse content to be prevalent and persistent before enforcement action can be taken is a positive one. It is welcome that Ofcom will have the opportunity to consider a range of factors.

Despite this, Labour—alongside the International Justice Mission—is still concerned about the inclusion of prevalence as a factor, owing to the difficulty in detecting newly produced CSEA content, especially livestreamed abuse. Amendments 35, 36, 39 and 40 seek to address that gap. Broadly, the amendments aim to capture the concern about the Bill’s current approach, which we feel limits its focus to the risk of harm faced by individuals in the UK. Rather, as we have discussed previously, the Bill should recognise the harm that UK nationals cause to people around the world, including children in the Philippines. The amendments specifically require Ofcom to consider the presence of relevant content, rather than its prevalence.

Amendment 37 would require Ofcom’s risk assessments to consider risks to adults and children through the production, publication and dissemination of illegal content—an issue that Labour has repeatedly raised. I believe we last mentioned it when we spoke to amendments to clause 8, so I will do my best to not repeat myself. That being said, we firmly believe it is important that video content, including livestreaming, is captured by the Bill. I remain unconvinced that the Bill as it stands goes far enough, so I urge the Minister to closely consider and support these amendments. The arguments that we and so many stakeholders have already made still stand.

[Kirsty Blackman]

I echo the sentiments that have been expressed by the shadow Minister, and thank her and her colleagues for tabling this amendment and giving voice to the numerous organisations that have been in touch with us about this matter. The Scottish National party is more than happy to support the amendment, which would make the Bill stronger and better, and would better enable Ofcom to take action when necessary.

[Chris Philp]

I understand the spirit behind these amendments, focusing on the word “presence” rather than “prevalence” in various places. It is worth keeping in mind that throughout the Bill we are requiring companies to implement proportionate systems and processes to protect their users from harm. Even in the case of the most harmful illegal content, we are not placing the duty on companies to remove every single piece of illegal content that has ever appeared online, because that is requesting the impossible. We are asking them to take reasonable and proportionate steps to create systems and processes to do so. It is important to frame the legally binding duties in that way that makes them realistically achievable.

As the shadow Minister said, amendments 35, 36, 39 and 40 would replace the word “prevalence” with “presence”. That would change Ofcom’s duty to enforce not just against content that was present in significant numbers—prevalent—but against a single instance, which would be enough to engage the clause.

We mutually understand the intention behind these amendments, but we think the significant powers to compel companies to adopt certain technology contained in section 103 should be engaged only where there is a reasonable level of risk. For example, if a single piece of content was present on a platform, if may not be reasonable or proportionate to force the company to adopt certain new technologies, where indeed they do not do so at the moment. The use of “prevalence” ensures that the powers are used where necessary.

It is clear—there is no debate—that in the circumstances where scanning technology is currently used, which includes on Facebook Messenger, there is enormous prevalence of material. To elaborate on a point I made in a previous discussion, anything that stops that detection happening would be unacceptable and, in the Government’s view, it would not be reasonable to lose the ability to detect huge numbers of images in the service of implementing encryption, because there is nothing more important than scanning against child sexual exploitation images.

However, we think adopting the amendment and replacing the word “prevalence” with “presence” would create an extremely sensitive trigger that would be engaged on almost every site, even tiny ones or where there was no significant risk, because a single example would be enough to trigger the amendment, as drafted. Although I understand the spirit of the amendment, it moves away from the concepts of proportionality and reasonableness in the systems and processes that the Bill seeks to deliver.

Amendment 37 seeks to widen the criteria that Ofcom must consider when deciding to use section 103 powers. It is important to ensure that Ofcom considers a wide range of factors, taking into account the harm occurring, but clause 104(2)(f) already requires Ofcom to consider

“the level of risk of harm to individuals in the United Kingdom presented by relevant content, and the severity of that harm”.

Therefore, the Bill already contains provision requiring Ofcom to take those matters into account, as it should, but the shadow Minister is right to draw attention to the issue.

Finally, amendment 38 seeks to amend clause 116 to require Ofcom to consider the risk of harm posed by individuals in the United Kingdom, in relation to adults and children in the UK or elsewhere, through the production, publication and dissemination of illegal content. In deciding whether to make a confirmation decision requiring the use of technology, it is important that Ofcom considers a wide range of factors. However, clause 116(6)(e) already proposes to require Ofcom to consider, in particular, the risk and severity of harm to individuals in the UK. That is clearly already in the Bill.

I hope that this analysis provides a basis for the shadow Minister to accept that the Bill, in this area, functions as required. I gently request that she withdraw her amendment.

--- Later in debate ---

[The Chair]

The Question is—

[Kirsty Blackman]

rose—

[The Chair]

I beg your pardon; I am trying to do too many things at once. I call Kirsty Blackman.

[Kirsty Blackman]

Thank you very much, Sir Roger. I do not envy you in this role, which cannot be easy, particularly with a Bill that is 190-odd clauses long.

[The Chair]

It goes with the job.

[Kirsty Blackman]

I have a quick question for the Minister about the timelines in relation to the guidance and the commitment that Ofcom gave to producing a road map before this coming summer. When is that guidance likely to be produced? Does that road map relate to the guidance in this clause, as well as the guidance in other clauses? If the Minister does not know the answer, I have no problem with receiving an answer at a later time. Does the road map include this guidance as well as other guidance that Ofcom may or may not be publishing at some point in the future?

[Chris Philp]

I welcome the cross-party support for the provisions set out in these important clauses. Clause 107 points out the requirement for Ofcom to publish guidance, which is extremely important. Clause 108 makes sure that it publishes an annual report. Clause 109 covers the interpretations.

The hon. Member for Aberdeen North asked the only question, about the contents of the Ofcom road map, which in evidence it committed to publishing before the summer. I cannot entirely speak for Ofcom, which is of course an independent body. In order to avoid me giving the Committee misleading information, the best thing is for officials at the Department for Digital, Culture, Media and Sport to liaise with Ofcom and ascertain what the exact contents of the road map will be, and we can report that back to the Committee by letter.

It will be fair to say that the Committee’s feeling—I invite hon. Members to intervene if I have got this wrong—is that the road map should be as comprehensive as possible. Ideally, it would lay out the intended plan to cover all the activities that Ofcom would have to undertake in order to make the Bill operational, and the more detail there is, and the more comprehensive the road map can be, the happier the Committee will be.

Officials will take that away, discuss it with Ofcom and we can revert with fuller information. Given that the timetable was to publish the road map prior to the summer, I hope that we are not going to have to wait very long before we see it. If Ofcom is not preparing it now, it will hopefully hear this discussion and, if necessary, expand the scope of the road map a little bit accordingly.

Question put and agreed to.

Clause 107 accordingly ordered to stand part of the Bill

Clauses 108 and 109 ordered to stand part of the Bill.

Clause 110

Provisional notice of contravention

Question proposed, That the clause stand part of the Bill.

--- Later in debate ---

[Alex Davies-Jones]

Labour welcomes this important clause, which lists the enforceable requirements. Failure to comply with those requirements can trigger enforcement action. However, the provisions could go further, so we urge the Minister to consider our important amendments.

Amendments 52 and 53 make it abundantly clear that more access to, and availability of, data and information about systems and processes would improve understanding of the online environment. We cannot rely solely on Ofcom to act as problems arise, when new issues could be spotted early by experts elsewhere. The entire regime depends on how bright a light we can shine into the black box of the tech companies, but only minimal data can be accessed.

The amendments would require Ofcom simply to produce a code of practice on access to data. We have already heard that without independent researchers accessing data on relevant harm, the platforms have no real accountability for how they tackle online harms. Civil society and researchers work hard to identify online harms from limited data sources, which can be taken away by the platforms if they choose. Labour feels that the Bill must require platforms, in a timely manner, to share data with pre-vetted independent researchers and academics. The EU’s Digital Services Act does that, so will the Minister confirm why such a provision is missing from this supposed world-leading Bill?

Clause 136 gives Ofcom two years to assess whether access to data is required, and it “may”, but not “must”, publish guidance on how its approach to data access might work. The process is far too slow and, ultimately, puts the UK behind the EU, whose legislation makes data access requests possible immediately. Amendment 52 would change the “may” to “must”, and would ultimately require Ofcom to explore how access to data works, not if it should happen in the first place.

[Kirsty Blackman]

Frances Haugen’s evidence highlighted quite how shadowy a significant number of the platforms are. Does the hon. Member agree that that hammers home the need for independent researchers to access as much detail as possible so that we can ensure that the Bill is working?

[Alex Davies-Jones]

I agree 100%. The testimony of Frances Haugen, the Facebook whistleblower, highlighted the fact that expert researchers and academics will need to examine the data and look at what is happening behind social media platforms if we are to ensure that the Bill is truly fit for purpose and world leading. That process should be carried out as quickly as possible, and Ofcom must also be encouraged to publish guidance on how access to data will work.

Ultimately, the amendments make a simple point: civil society and researchers should be able to access data, so why will the Minister not let them? The Bill should empower independently verified researchers and civil society to request tech companies’ data. Ofcom should be required to publish guidance as soon as possible —within months, not years—on how data may be accessed. That safety check would hold companies to account and make the internet a safer and less divisive space for everyone.

The process would not be hard or commercially ruinous, as the platforms claim. The EU has already implemented it through its Digital Services Act, which opens up the secrets of tech companies’ data to Governments, academia and civil society in order to protect internet users. If we do not have that data, researchers based in the EU will be ahead of those in the UK. Without more insight to enable policymaking, quality research and harm analysis, regulatory intervention in the UK will stagnate. What is more, without such data, we will not know Instagram’s true impact on teen mental health, nor the reality of violence against women and girls online or the risks to our national security.

We propose amending the Bill to accelerate data sharing provisions while mandating Ofcom to produce guidance on how civil society and researchers can access data, not just on whether they should. As I said, that should happen within months, not years. The provisions should be followed by a code of practice, as outlined in the amendment, to ensure that platforms do not duck and dive in their adherence to transparency requirements. A code of practice would help to standardise data sharing in a way that serves platforms and researchers.

The changes would mean that tech companies can no longer hide in the shadows. As Frances Haugen said of the platforms in her evidence a few weeks ago:

“The idea that they have worked in close co-operation with researchers is a farce. The only way that they are going to give us even the most basic data that we need to keep ourselves safe is if it is mandated in the Bill. We need to not wait two years after the Bill passes”.––[Official Report, Online Safety Public Bill Committee, 26 May 2022; c. 188, Q320.]

--- Later in debate ---

[Chris Philp]

This data is a little different—the two domains do not directly correspond. In the health area, there has been litigation—an artificial intelligence company is currently engaged in litigation with an NHS hospital trust about a purported breach of patient data rules—so even in that long-established area, there is uncertainty and recent, or perhaps even current, litigation.

We are asking for the report to be done to ensure that those important issues are properly thought through. Once they are, Ofcom has the power under clause 136 to lay down guidance on providing access for independent researchers to do their work.

[Kirsty Blackman]

The Minister has committed to Ofcom being fully resourced to do what it needs to do under the Bill, but he has spoken about time constraints. If Ofcom were to receive 25,000 risk assessments, for example, there simply would not be enough people to go through them. Does he agree that, in cases in which Ofcom is struggling to manage the volume of data and to do the level of assessment required, it may be helpful to augment that work with the use of independent researchers? I am not asking him to commit to that, but to consider the benefits.

[Chris Philp]

Yes, I would agree that bona fide academic independent researchers do have something to offer and to add in this area. The more we have highly intelligent, experienced and creative people looking at a particular problem or issue, the more likely we are to get a good and well-informed result. They may have perspectives that Ofcom does not. I agree that, in principle, independent researchers can add a great deal, but we need to ensure that we get that set up in a thoughtful and proper way. I understand the desire to get it done quickly, but it is important to take the time to do it not just quickly, but right. It is an area that does not exist already—at the moment, there is no concept of independent researchers getting access to the innards of social media companies’ data vaults—so we need to make sure that it is done in the right way, which is why it is structured as it is. I ask the Committee to stick with the drafting, whereby there will be a report and then Ofcom will have the power. I hope we end up in the same place—well, the same place, but a better place. The process may be slightly slower, but we may also end up in a better place for the consideration and thought that will have to be given.

--- Later in debate ---

Oversight is required to ensure that human resources processes clearly identify the role and provide content descriptions, as well as information on possible occupational hazards. Currently, the conditions of the work are unregulated and rely on the business relationship between two parties focused on the bottom line. Platforms do not release any due diligence on the employment conditions of those contractors, if they conduct it at all. If there is to be any meaningful oversight of the risks inherent in the content moderation supply chain, it is imperative to mandate transparency around the conditions for content moderators in contracted entities. As long as that relationship is self-regulated, the wellness of human moderators will be at risk. That is why we urge the Minister to support this important amendment and new clause: there is a human element to all this. We urge him to do the right thing.

[Kirsty Blackman]

I thank the hon. Member for Pontypridd for laying out her case in some detail, though nowhere near the level of detail that these people have to experience while providing moderation. She has given a very good explanation of why she is asking for the amendment and new clause to be included in the Bill. Concerns are consistently being raised, particularly by the Labour party, about the impact on the staff members who have to deal with this content. I do not think the significance of this issue for those individuals can be overstated. If we intend the Bill to have the maximum potential impact and reduce harm to the highest number of people possible, it makes eminent sense to accept this amendment and new clause.

There is a comparison with other areas in which we place similar requirements on other companies. The Government require companies that provide annual reports to undertake an assessment in those reports of whether their supply chain uses child labour or unpaid labour, or whether their factories are safe for people to work in—if they are making clothes, for example. It would not be an overly onerous request if we were to widen those requirements to take account of the fact that so many of these social media companies are subjecting individuals to trauma that results in them experiencing PTSD and having to go through a lengthy recovery process, if they ever recover. We have comparable legislation, and that is not too much for us to ask. Unpaid labour, or people being paid very little in other countries, is not that different from what social media companies are requiring of their moderators, particularly those working outside the UK and the US in countries where there are less stringent rules on working conditions. I cannot see a reason for the Minister to reject the provision of this additional safety for employees who are doing an incredibly important job that we need them to be doing, in circumstances where their employer is not taking any account of their wellbeing.

[Barbara Keeley]

As my hon. Friend the Member for Pontypridd has pointed out, there is little or no transparency about one of the most critical ways in which platforms tackle harms. Human moderators are on the frontline of protecting children and adults from harmful content. They must be well resourced, trained and supported in order to fulfil that function, or the success of the Bill’s aims will be severely undermined.

I find it shocking that platforms offer so little data on human moderation, either because they refuse to publish it or because they do not know it. For example, in evidence to the Home Affairs Committee, William McCants from YouTube could not give precise statistics for its moderator team after being given six days’ notice to find the figure, because many moderators were employed or operated under third-party auspices. For YouTube’s global counter-terrorism lead to be unaware of the detail of how the platform is protecting its users from illegal content is shocking, but it is not uncommon.

In evidence to this Committee, Meta’s Richard Earley was asked how many of Meta’s 40,000 human moderators were outsourced to remove illegal content and disinformation from the platform. My hon. Friend the Member for Pontypridd said:

“You do not have the figures, so you cannot tell me.”

Richard Earley replied:

“I haven’t, no, but I will be happy to let you know afterwards in our written submission.”

Today, Meta submitted its written evidence to the Committee. It included no reference to human content moderators, despite its promise.

The account that my hon. Friend gave just now shows why new clause 11 is so necessary. Meta’s representative told this Committee in evidence:

“Everyone who is involved in reviewing content at Meta goes through an extremely lengthy training process that lasts multiple weeks, covering not just our community standards in total but also the specific area they are focusing on, such as violence and incitement.”––[Official Report, Online Safety Public Bill Committee, 24 May 2022; c. 45, Q76.]

But now we know from whistleblowers such as Daniel, whose case my hon. Friend described, that that is untrue. What is happening to Daniel and the other human moderators is deeply concerning. There are powerful examples of the devastating emotional impact that can occur because human moderators are not monitored, trained and supported.

There are risks of platforms shirking responsibility when they outsource moderation to third parties. Stakeholders have raised concerns that a regulated company could argue that an element of its service is not in the scope of the regulator because it is part of a supply chain. We will return to that issue when we debate new clause 13, which seeks to ensure enforcement of liability for supply chain failures that amount to a breach of one of the specified duties.

Platforms, in particular those supporting user-to-user generated content, employ those services from third parties. Yesterday, I met Danny Stone, the chief executive of the Antisemitism Policy Trust, who described the problem of antisemitic GIFs. Twitter would say, “We don’t supply GIFs. The responsibility is with GIPHY.” GIPHY, as part of the supply chain, would say, “We are not a user-to-user platform.” If someone searched Google for antisemitic GIFs, the results would contain multiple entries saying, “Antisemitic GIFs—get the best GIFs on GIPHY. Explore and share the best antisemitic GIFs.”

One can well imagine a scenario in which a company captured by the regulatory regime established by the Bill argues that an element of its service is not within the ambit of the regulator because it is part of a supply chain presented by, but not necessarily the responsibility of, the regulated service. The contracted element, which I have just described by reference to Twitter and GIPHY, supported by an entirely separate company, would argue that it was providing a business-to-business service that is not user-generated content but content designed and delivered at arm’s length and provided to the user-to-user service to deploy for its users.

I suggest that dealing with this issue would involve a timely, costly and unhelpful legal process during which systems were not being effectively regulated—the same may apply in relation to moderators and what my hon. Friend the Member for Pontypridd described; there are a number of lawsuits involved in Daniel’s case—and complex contract law was invoked.

We recognise in UK legislation that there are concerns and issues surrounding supply chains. Under the Bribery Act 2010, for example, a company is liable if anyone performing services for or on the company’s behalf is found culpable for specific actions. These issues on supply chain liability must be resolved if the Bill is to fulfil its aim of protecting adults and children from harm.