Crime and Policing Bill

Lord Clement-Jones Excerpts

Wednesday 18th March 2026 Lords Chamber

[Lord Russell of Liverpool (CB)]

My Lords, I will be brief. I entirely support the noble Baroness, Lady Kidron, on all her amendments. What I would say to the Government about their own amendment is that I have just had what I suppose is the privilege—although it sometimes seemed quite lengthy—of being a member of the Secondary Legislation Scrutiny Committee, and I can tell noble Lords that the quality of much secondary legislation is lamentable, varying by department. A lack of preparation, of any Explanatory Memorandum explaining anything relevant, and of any impact assessment whatsoever, is extremely frequent. In the last year, we have had several secondary instruments relating directly to the Online Safety Act, none of which has been particularly impressive, and some of which have been debated on the Floor of this House—my noble friend Lord Clement-Jones will be well aware of that. We have expressed our displeasure at the way in which this has been brought forward and explained.

All of us on the Cross Benches remember the late, lamented Lord Igor Judge. What he would think about a Government of this political hue bringing forward Henry VIII powers, to the power of 10, I cannot even imagine. If he is up there, he will be smiling wryly but he will not be impressed.

My only other point is rather strange. His Majesty’s occasionally loyal Opposition were extremely good at bringing in a variety of legislation which had a lot of Henry VIII powers. They have suddenly had a conversion on the road to Damascus, for which we should all be grateful. However, we need to think very carefully before we give the Government Henry VIII powers in an area as sensitive as this, and that is doing much harm as we speak.

[Lord Clement-Jones (LD)]

My Lords, I express from these Benches our very strong support for these comprehensive amendments tabled by the noble Baroness, Lady Kidron, which she has characteristically introduced so well and to which so many noble Lords have spoken so eloquently in support. I also want to express our concerns regarding the Government’s proposed alternative, Amendment 429B.

In this group, we confront digital harm that is not incidental but engineered by design. AI chatbots are no longer a futuristic curiosity but deeply embedded the lives of our children. They are designed not merely as tools but as confidantes, mentors, companions and, in some cases, explicit romantic partners. Their anthropomorphic features create dangerous emotional dependency. Without statutory safeguards, these bots can provide explicit information on how to self-harm. This is not a flaw but a design feature that drives engagement, and we cannot allow the generative power of AI to become a generator of despair.

We are not debating theoretical risks, as many noble Lords have said today. We are debating the forces that led to the tragic deaths of Sewell Setzer III, mentioned by a number of noble Lords, and Adam Raine, in the United States. Their families are pursuing legal action in the US on the basis that deceptively designed, inadequately safeguarded chatbots can be treated as defective products, and that developers should bear full legal liability when systems encourage, facilitate or fail to interrupt a user’s path to suicide.

I welcome the Government’s admission that a legal loophole exists in the UK. However, their proposed remedy, Amendment 429B, gives us a choice between the clarity of primary legislation through the amendments tabled by the noble Baroness, Lady Kidron, and the convenience of the Executive. In contrast, the noble Baroness’s amendments provide clarity and embed safety duties in the Bill. Like my noble friend, I highlight Amendment 433, which deals with targeting the engineered features that keep children hooked. We know that bots guilt-trip users who try to end conversations. For a child, this is not a user interface quirk; it is emotional manipulation. These amendments would prohibit such coercive engagement techniques and, crucially, require bots to signpost users to help when asked about health, suicide or self-harm.

The primary legislation route offered by these amendments is the only fully viable and responsible path. If the noble Baroness wants to test the opinion of the House, we will support her in the Lobby. Should we be unable to secure her amendments, we would need to take a view on Amendment 429B. Four specific binding assurances would be required before we could consider supporting it; without them, it is nothing but a dangerous blank cheque. As changing these sections effectively rewrites the criminal threshold of the Online Safety Act, the Government must commit to the equivalent of the super-affirmative procedure for all significant policy choices, including amendments to core definitions or the expansion of duties beyond priority legal content. Standard procedures will not give this House the scrutiny needed.

Regarding mandatory supply chain transparency, we need a firm commitment that regulations will include a statutory mandate for providers to document and share their technical blueprints with Ofcom. Without this, the regulator cannot do its job. The Minister must confirm that the power will be used to tackle the issues raised by subsections (6) and (7) of Section 192 of the Online Safety Act, ensuring that chatbots cannot evade regulation simply because they lack a human mens rea. A bot does not intend harm, but it can be designed to cause it. The Minister must commit that any new regulations will explicitly disapply the requirement to prove human intent for AI-generated content. Regulations must define control across the entire AI supply chain so that accountability is not lost in a black box.

Finally, we would require a clear assurance that this power will not be used to alter the legal position of services that are not AI services. The scope of Amendment 429B must not drift beyond its stated purpose. If the Government are serious when they say that no platform gets a free pass, that must apply equally to generative AI models that, as we speak, are reshaping the childhoods of so many of our citizens. Safety by design must be the price of entry into the UK market, not an aspiration deferred to secondary legislation.

[Lord Davies of Gower (Con)]

My Lords, both the noble Baroness, Lady Kidron, and the Government are trying to solve the same problem, but with different solutions. I have to say that I find both solutions wanting. I feel that the position I see solidifying in the House is that we must accept either the Government’s amendment or the noble Baroness’s amendment, that something must be done and that this is binary, and I am not sure that I can accept that. I commend the aims of the noble Baroness’s campaign and I hope that we can find some common ground, but I have a number of questions about her amendments.

The approach that the noble Baroness has taken in her Amendments 422D and 433 is to create criminal offences for a person to create or even supply an AI chatbot that produces a selection of prohibited content. That would place criminal penalties on individuals who are involved in the creation of such a chatbot. The Online Safety Act applies civil penalties when companies violate the regulations: the companies are fined by Ofcom if they allow prohibited content to be published on their platform. These amendments would apply criminal sanctions punishable by up to five years’ imprisonment, but who is liable for these criminal penalties? Is it the software engineer who developed the chatbot? Is it the employee who presses the “publish” button that brings the AI online? Is it the senior management team who oversaw it, or the investors who commissioned it? Is it the CEO of the whole company? It is not clear to me, and it would be useful if the noble Baroness could clear that up.

The offence of supplying such a chatbot might also be problematic. If an AI chatbot app is listed on the App Store, and the AI could in some way be used by a person in the manner described, should Apple be criminally liable for that? Similarly, I have a few concerns about the risk assessment that the amendment would require AI providers to undertake. For example, proposed new subsection (5)(e) would require that a provider

“assesses the risks to equality of treatment of individuals”.

I question whether it is the Government’s role to mandate the target audience of a business product. It is worrying enough to believe that it is meritocratic to mandate quotas within organisations, but it is quite another stance to say that the very reason for a business’s existence, its output, should be directed by legislation.

As I have made clear, we do not oppose the noble Baroness’s objective of addressing the harms of AI with this amendment, but simply saying that there is a problem and that doing anything is better than doing nothing, irrespective of the problems with that something, is not a proper way to legislate. It is a recipe, I suggest, for bad law. However, I understand that the noble Baroness is trying to make the Government take action. It is up to the Government to come to Parliament with a sufficient solution. The Minister may try to say that the Government do have a sufficient solution in government Amendment 429B, but that would be wholly incorrect. I am quite shocked that the Minister has even considered bringing this amendment to the House, and I can only imagine what he might have said about it if he were standing where I am now.

Amendment 429B grants the Secretary of State sweeping Henry VIII powers to amend the entire Online Safety Act for the purpose of mitigating harms presented by AI-generated content. I am sure that the noble Lord, Lord Russell of Liverpool, would be delighted to hear me say that this is an egregious attempt by an overreaching Government to exploit a serious issue to centralise power in the hands of the Secretary of State. It is almost unprecedented to grant a Minister the ability to amend an entire Act of Parliament. With this amendment, the Government are doing away with every bit of lip service they have paid to the importance of parliamentary scrutiny or their democratic mandate. The amendment would give not only this Government, who have made it clear that they are very happy running a centralised state with digital IDs, but every future Government the ability to amend online regulations and curtail the freedoms of providers. Indeed, a future Reform Government might go in the opposite direction and remove all regulations on AI. The noble Lord should reflect on that, too.

I ask the Minister to imagine that the glove was on the other hand: that he was standing at this Dispatch Box and I was the Minister proposing to give my Government these powers. There is no way that he would support such sweeping powers to amend an Act of Parliament by ministerial fiat. This is the Henry VIII power to end all Henry VIII powers. It cannot be allowed to make its way into the Bill and, when it is called, I will take pleasure in opposing it in the Lobbies.