To ask His Majesty’s Government what assessment they have made of the case for a cross-sector AI regulation bill.

My Lords, it is a pleasure to bring this QSD to the Grand Committee for debate. In doing so, I declare my interests as set out in the register, variously as adviser to the Crown Estate, Endava plc and Simmons and Simmons LLP. I thank all noble Lords who have signed up to speak, and I look forward to their quick-fire contributions.

I have been given eight minutes to open this debate, but I can do it in one word: is there a case for a cross-sector AI Bill? Yes. To expand somewhat on that, the Prime Minister has described AI as:

“The defining opportunity of our generation”,

yet the Government are largely taking a wait-and-see, voluntary and so-called domain-specific approach. I am not sure that wait and see is ever an optimal approach to any issue, particularly one as significant as AI. But do not listen to me; let us consider, on its own merits, how the Government’s approach is going. Harms are unaddressed. Young people are not getting shortlisted for jobs, without even knowing that it is AI that is kicking them out of the process; even if they knew, there would be little, if any, redress at this time. Job seekers, benefit claimants, teachers and teenagers are all suffering the harms of AI that are currently unaddressed. Similarly, vast opportunities are being unoptimised for the UK. Wait and see has really led to partial, piecemeal and voluntary action.

Why have the Government taken this approach? Let us take just two of the elements that are offered. The first is that it is too soon to legislate; they will stymie, stifle or stop progress. Not a bit of it—we know exactly what we need to do to put in place the right-sized legislation and regulation. The second is a falsehood that recurs with tedious inevitability: you can either have regulation or innovation, but you cannot have both. We all know bad regulation—there is a deal of it about—but that does not mean for one second that regulation, of itself, is bad. Right-sized and right-touch regulation is good for citizens, creatives, consumers, innovators and investors alike. We know how to do this. Just look back to the telco regulation of some decades ago. Was it stopping an industry? Far from it: it was a key enabler of Great British telco business.

What do we need to do? What anyone needs are clarity, consistency and coherence. This is currently not the case. We have such an opportunity because we have the great good fortune of common law, which is agile, adaptable and ideal for the task at hand. We need to bring forward principles-based, outcomes-focused, inputs-understood legislation. Those principles are trust and transparency; inclusion and innovation; interoperability and international outlook; accountability, assurance and accessibility—and we need to put those principles on a statutory footing.

We need to look to AI responsible officers across business. We need labelling to address the IP challenge, which currently continues to be kicked down the road, with creatives’ work taken—unremunerated, unrespected and unconsented to. Crucially, we need public engagement across this if we are going to enable all the opportunities from AI. We know what to do. We have the legislative capability, knowledge and history of how to do this.

Do not just listen to me. Let us take the great Ada Lovelace Institute, which variously describes the Government’s approach as chronically insufficient and, in a glorious understatement,

“increasingly out of step with public attitudes”.

If the Government want growth and public trust in AI and optimal outcomes across the whole of society, we need to end this wait-and-see, voluntary, partial, piecemeal approach. For economic, social, democratic and psychological reasons, we need a cross-sector, principles-based, outcomes-focused AI regulation Bill now. We need this for citizens, creatives, consumers, innovators and investors all to be enabled and empowered to say full-throatedly together, “Our data, our decisions, our AI futures”—our human-led, inclusive AI futures.

My Lords, my first duty is to thank the noble Lord, Lord Holmes, for that introduction. He and I are members of the Lords Communications and Digital Committee. We have both benefited from discussions and hearings about AI held by the Select Committee under the shrewd guidance of our chair, the noble Baroness, Lady Keeley.

Our briefings seemed to come from two directions. First, the tech sector sees the onward march to artificial intelligence to be a unique opportunity to achieve unprecedented leaps in productivity, problem-solving and efficiency—the very combination of which successive Governments have been seeking since the great financial crash of 2008. There are counterarguments, coming mainly from the creative industries and from people such as the most reverend Primate the Archbishop of Canterbury and the Pope, who see the risks and dangers associated with AI as enough to give us pause before we decide on the direction to take.

If the AI revolution is already on the way—I believe it is—our first task is to ensure that our machinery of government is fit for purpose. The noble Lord, Lord Holmes, suggested a cross-sector AI Bill. I would go further and establish a cross-party, pre-legislative scrutiny committee of both Houses to take evidence on the proposals in such a Bill. I was on the committee for the Communications Act 2003. It was one of the great experiences of my life, in exchanging and hearing ideas.

Something else that has been suggested is the setting up of a COBRA-type committee. I also served on COBRA in the past. AI calls for the kind of committee that is able to look forward, anticipate problems and advise on what we are doing.

These would be two useful first steps in making the AI revolution less painful and destructive to the lives of working people than other tech innovations have in the past.

My Lords, I too thank the noble Lord, Lord Holmes, for securing this debate. I will echo some of his themes.

Cross-sectoral regulation will come. The question is simply: when? This week, the Trump Administration issued an executive order compelling new AI frontier models to be submitted for review. All general-purpose technologies require regulation—electricity, rail and nuclear. Like nuclear, AI is both a general-purpose technology and a dual-use one. But unlike early nuclear, AI innovation is occurring in private companies and not in government labs, which partly explains the regulatory squeamishness. The US fears falling behind China; the UK fears impeding growth and points to Europe, where regulation has damaged innovation. As the noble Lord, Lord McNally, said, into this debate stepped the Pope, the most reverend Primate the Archbishop of Canterbury and a host of lesser luminaries, including the head of MI5, the founders of Anthropic and military chiefs—all agitating for action both locally and nationally.

Regulation is not antithetical to growth and to realising the upsides of AI. The Government are delivering cross-sectoral legislation on cybersecurity but have hesitated on a comparable approach to AI safety. I invite the Minister, when summing up, to share when the Government will launch their consultation on AI regulation. Will they consider regulation at the point of development and not simply at the point of use? Will they mandate pre-deployment testing of high-risk AI, as the US has now done? Do they accept the case for an AI regulatory oversight body?

These are robust questions for my own Government. Yet these issues are too big for us to play politics with. I note that I am to be followed by the noble Lord, Lord Harper, who is unquestionably the master of robust challenge. Yet his Government also wrestled with the growth versus regulation dilemma; Rishi Sunak counselled caution as recently as last week about us being a first mover. I welcome the call of the noble Lord, Lord McNally, for a cross-party approach. The nation and the world are looking to us for actions that match our moment. This requires an inclusive discussion, debate and direction. It is time to act.

My Lords, I thank my noble friend Lord Holmes for securing this debate, which has given us the opportunity, albeit briefly, to talk about this subject. I also thank the noble Baroness, Lady Alexander, for what I think was intended as a kind remark.

I support a principles-based approach, which is not surprising since that was the approach set out by the Government of which I was a member. A number of principles should be adopted by existing regulators—I do not support having a single AI regulator—balanced against their existing duties on growth and innovation. This is important because, while there are absolutely risks with AI, it will be one of the biggest potential drivers of growth and innovation if we get it right. We need to balance those two things.

I also want to focus on the thing that we are doing well—the AI Security Institute, which we set up and the current Government have retained. I was listening to Matt Clifford the other day, an apolitical official who is experienced in this sector. He was brought in by Rishi Sunak and retained by the present Prime Minister. He made the point that this is world leading, enabling us to robustly test AI models and approaches to make sure that we deal with the risks. It is very successful at enabling us to be a global leader. We are never going to be like the United States, the centre of all these things, but outside of the US and China we are a leading player in this area.

I support a principles-based approach and a relatively short piece of legislation that would give the duty to existing regulators to adopt those principles. However, we need to move at pace and not do what happened with the Online Safety Act—still talking about the legislation in seven years’ time, when, frankly, the world will have moved on at pace.

My Lords, I too congratulate the noble Lord, Lord Holmes of Richmond, on his commitment to AI regulation and on securing today’s QSD. In the time available, I will follow the noble Lord, Lord Harper, and concentrate on AISI, the AI Security Institute.

When Claude Mythos Preview was released by Anthropic to a select group of launch partners a few weeks ago, AISI was the only non-American government organisation to receive access to the model for safety testing. However, should the UK Government in three years’ time have the same view of AI governance as the current US President, AISI’s future would be under threat. The Government could decide overnight to repurpose it or even shut it down altogether.

I therefore proposed a Private Member’s Bill, not drawn in the ballot, which would establish AISI as an independent statutory body. It would make provision for its functions in relation to the testing and evaluation of advanced AI systems, including pre-deployment and post-deployment assessment of risks, and for the institute to give advice to the Government and Parliament. Putting AISI on a statutory footing, beyond its current status as a unit within DSIT, would signal to the international community that the UK’s commitment to AI safety was permanent.

We cannot wait: Anthropic is predicting that there is a greater than 50% probability that AI self-recursive improvement will be achieved by 2028. If AISI has become an independent statutory body by then, it will be able to provide regulatory advice on whether AI designed by AI is safe to use or whether it should be banned in the UK. Expert-driven regulation based on advice from an independent AISI, using Henry VIII powers if required, is the best option for AI safety in the UK. I urge the Minister and the Government to allocate some bandwidth to thinking about this critical issue.

My Lords, I too thank my noble friend Lord Holmes for bringing this debate. The fact that so many of us wish to speak today shows that this really needs to be properly debated as a society, not just in a series of two-minute soundbites.

I think we all agree that good regulation and economic growth are not a trade-off—they go hand in hand. In order to get good regulation, you have to go through quite detailed and thoughtful work. My question—I have asked the Minister this before in a Question—is: why are the Government not initiating a commission along the lines of the Warnock committee that looked at human embryology and fertilisation in the 1980s?

This single hour shows that there are really important moral, social and economic issues about AI that we need to think through. With the greatest of respect to my noble friend beside me, I genuinely do not think that individual regulators on their own can sort that out. We have to have a proper societal debate about this, and we have the democratic tools to do it. That is not a hurried piece of emergency legislation; it is a proper commission, a Green Paper and a White Paper. It is about building the societal trust that we need for this ground-breaking regulation.

I urge the Government to think carefully, because at the moment we are doing nothing, and that is really dangerous. Nothing takes us to the same place that we are in with social media, which is a huge amount of harm done, no action and then calls to ban the technology. Banning the technology is a really bad idea; regulating it well is where we should be going.

I too congratulate the noble Lord, Lord Holmes, on securing this debate. Given the pace of AI development, it could not be more timely. As Pope Leo said in his recent encyclical:

“Each generation inherits the task of shaping its own era, of guiding history to become a place where the dignity of every person is safeguarded, justice is promoted and fraternity is made possible”.

Regulation does not of necessity stifle innovation. As David Epstein argues in his recent book Inside the Box, creativity, innovation and problem-solving are often improved by constraints rather than exhortations to “think outside the box”. The need for regulation in this space, however, goes well beyond this. One cannot divorce current technological advances from the moral framework which underlies them. Every advance reflects a moral vision and an ideological bias. There is—in some quarters of the AI industry at least—a dehumanising conviction that evolution drives us towards perfection. AI, it is argued, will be better than us at many things, so why not harvest the best of us and move on? As recent debates about social media use by under-16s demonstrate, large technology companies motivated exclusively by profit do not necessarily have our best interests at heart. Such utilitarianism is not a basis for a human ethical framework.

Regulation is ultimately about the restraint of such power for the common good. There are underlying systems which power many of the applications with which we are familiar, and it is these that need to be regulated. As many noble Lords have said, it is insufficient to do this piecemeal by sector when the risks that experts warn about are not sector specific. They are systemic, and potentially catastrophic. Noble Lords in this debate have already suggested many methodologies for bringing that into practice. I urge His Majesty’s Government, in collaboration with other nations, to introduce guardrails at the point of development.

My Lords, I declare my interests as deputy chairman of the Telegraph Media Group and chairman of the News Media Association. The vast, unaccountable technology companies would have us believe that any attempt at regulation in this area would stifle their growth and damage the economy. That is nonsense, of course. The truth is that legislation, even of the most modest sort, is long overdue to protect public safety, ensure transparency, tame untrammelled market power and protect content and intellectual property.

I have two points. First, there is already legislation on the statute book that, as a starting point, simply needs to be enforced. Rather than trying to rip the guts out of the CMA, the Government should actively direct it to encourage greater competition in digital markets. Google’s announcement yesterday that it will give publishers greater control over use of their work globally, after the CMA introduces binding conduct requirements on search, not only demonstrates that big tech will not flee if we introduce proportionate regulation but shows that the UK has an opportunity to lead the way.

Secondly, the Government should act now to ensure that existing copyright law is meaningfully enforced. For news publishers, indeed all content publishers, whose content is routinely scraped—or stolen—without consent or remuneration, simple transparency over the identity and purpose of the “crawlers” accessing that content would provide the basic information required to assert their rights and be rewarded for them. It is not a radical demand and falls well short of proper regulation, but it would be a start as a practical measure to make existing law effective. Yet even this has been kicked into the long grass, despite consistent demands from this House for action.

We should not accept political paralysis on this vital issue. AI can be a great force for good but, to serve all of society, laws must be enforced and fresh regulation put in place at speed. Unless we act now, history will taunt us with its most feared words: “too late”.

My Lords, the noble Lord, Lord Holmes, has done us all a favour by facilitating this discourse, but it should not be necessary. In the 2024 King’s Speech, the Government committed to placing binding regulations on the developers of the most powerful AI models. Almost two years later, we still await legislation to fulfil this commitment. I therefore hope that my noble friend will be able to explain why that important policy has been dropped in favour of a weaker approach through targeted regulation at the point of use.

This ignores the warnings by experts developing superintelligent AI and leaves a major regulatory gap because recent models, such as Claude Mythos, have exposed the serious risks that originate at the point of development. On this occasion, the developers decided not to release the model to the public owing to its dangerous capabilities, but a more blindly profit-driven company or hostile actor in the future may not, and we cannot afford to rely on private companies doing the right thing to avoid national and global security threats.

It has been said that the pace of technological change is greater today than it has ever been, but never again will it be as slow as it is today. Companies are investing in what is known as recursive self-improvement, which means AIs autonomously developing more AIs. That would be the point at which we as humans lose control—a truly frightening prospect. MI5 director-general Ken McCallum recently commented that

“Artificial intelligence may never ‘mean’ us harm. But it would be reckless to ignore the potential for it to cause harm”.

I declare an interest of sorts, because I am one of more than 100 parliamentarians in both Houses who have joined ControlAI’s campaign on raising awareness of superintelligence and calling for binding regulation on the most powerful models. British mathematicians and scientists have done much to advance Al technology and the leading ones—the Nobel laureates and “godfathers of AI”—are loudest in their calls for proper regulation. It is past time that the Government took note and acted decisively.

My Lords, I, too, thank the noble Lord, Lord Holmes of Richmond, for securing this important debate. Innovation and technology have brought solutions to everyday problems. Britain can be proud of the fact that it started the Industrial Revolution, but we are now in the fast-moving AI revolution. Either we regulate or we stagnate.

It was my Bill that, in 1997, established Britain’s first ever comprehensive DNA database. However, without DNA regulation, we would not have experienced that positive DNA revolution. As a result, we became the first nation in the world to embrace this ground-breaking technology. Nearly 30 years later, the DNA database has positively transformed the effectiveness of medicine, forensic science and research science. Some 175 of the 195 nations in the world now utilise DNA profiles.

There was no mention of an AI regulation Bill in the King’s Speech. Meanwhile, AI is being monitored in pieces through the Online Safety Act and a patchwork of policies in various sectors. This is not the same as a coherent framework for the development, deployment and use of AI. It seems to me that clear regulation on the basis of established common-law principles is not the enemy of innovation. Can the Minister please indicate when we are likely to see an AI regulation Bill?

The Cyber Security and Resilience (Network and Information Systems) Bill is long overdue. In recent months, we have seen airports such as Heathrow, retail chains such as Marks & Spencer, and large manufacturers such as Jaguar Land Rover all be immobilised by hacking. Two days ago, President Trump created a framework, or clearing house, for federal government to vet powerful new AI models before they are released. This stemmed from Anthropic’s warning that it has developed an AI tool, Claude Mythos, which is an advanced cyber security threat. Will the Minister follow this policy but go one step further than the Americans and make a UK AI clearing house mandatory, not voluntary?

For 10 years, I had the privilege of being the vice-president of the British Board of Film Classification. We regulated films and videos from all over the world, including the “Star Wars” epics. Unless we introduce effective AI regulation now, AI could become less Luke Skywalker and more Darth Vader.

My thanks go to my noble friend Lord Holmes for securing this debate.

The last thing that one of the few sectors in the British economy that is growing needs is more regulation. The example of Anthropic referring Mythos to the AI Security Institute in order to do a safety evaluation shows that AI companies have a commercial incentive to prioritise safety. That is not an argument for a cross-sector regulator. I am concerned that more red tape will favour well-resourced incumbents who can afford compliance departments and penalise scrappy upstarts.

The UK is currently third in the world in the AI arms race. How long will it retain that position if we introduce heavy-handed regulation to the UK, which would put us at a competitive disadvantage? I urge the Government not to think about creating a cross-sector AI regulator but, instead, to think about the transnational regulation of AI and to engage with some of our trading partners on developing a transnational regulatory framework that could be built into trade agreements. The natural trading partner to begin that conversation with, if you are not having it already, is the United States, which is the leading developer of AI in the world. I would prefer a transnational regulatory system for AI to be based on the principles of the first amendment and developed in partnership with the United States—perhaps with Switzerland, too, as another non-EU European country—rather than favouring the model in the EU’s AI safety Act, which, like the EU’s Digital Services Act, prioritises safety over liberty and innovation to far too great an extent.

I, too, thank the noble Lord, Lord Holmes, for this debate and for my first opportunity to speak in Grand Committee. I want to dwell on two connected examples that demonstrate both the potential benefits and dangers of using AI on the battlefield, including why defence needs a bespoke approach. In doing so, I acknowledge the excellent research of Katrina Manson and Kevin T Baker on the US development of AI in warfare under the auspices of Project Maven.

In 2011, a friendly fire incident took place, killing US marines fighting in Afghanistan. The troops were mistakenly identified as the Taliban and targeted by a US drone. However, the way in which the marines were arranged, prone on the ground, was not the way the Taliban fought. This motivated senior US officers to push for the increased use of AI in order to prevent human error and was a key factor in the journey that the US military took in developing Project Maven.

In March this year, US weapons hit a primary school in Minab in southern Iran, killing more than 150 people, mostly schoolgirls. The knee-jerk reaction was to blame AI, but people had built a system that was fast enough to make a failure to update the target database lethal. Nobody searched the database to check that the target was legitimate because, with the Project Maven system making 1,000 decisions an hour, nobody could.

I offer as a conclusion that the use of AI in warfare does not lend itself to cross-sector regulation; in fact, I am much persuaded by the approach suggested by the noble Baroness, Lady Harding, more generally. In the defence context, a question that remains unanswered is whether AI is a tool, with humans at the helm, or a killer in its own right. The ethical implications relate to the tension between reducing adherence to the law of armed conflict and risking unintended outcomes, or losing in warfare because the enemy has no qualms about those risks. There is no time now to dwell further on this important moral challenge, but perhaps we should debate it further in the House in future.

My Lords, I also thank the noble Lord, Lord Holmes, for obtaining this debate, and more so for his singular and determined leadership and focus on this issue. The need is becoming even more urgent for the Government to end their current policy of “just wait and see”. They should go for an overall scheme and unified set of principles, preferably with a single regulator.

I agree with the noble Baroness, Lady Harding, that regulation is needed to underpin growth and development. Great though the common law is, on its own it is not enough. I illustrate this with reference to the legal sector, which, to the advantage of the United Kingdom, has a significant international lead at present. The sector is making major investment in AI, with real change to the way in which lawyers have trained and work is under way. This is now having a marked effect on the business of the courts and arbitration, and to the competitiveness of the industries it serves, particularly our leading financial services sector.

The issues that arise in the legal sector are many. To list but a few, they include confidentiality, transparency, integrity, copyright, automatic decision-making, protection from bias and discrimination, equal treatment, training, storage of data and ensuring the competitive market for AI. These are exactly the same as exist elsewhere and should be guided by a single set of principles with sectoral adjustment where necessary. Such a uniform set of principles is, in my view, essential to the maintenance of the leading edge of the legal sector in its international position in what is a fiercely competitive international market. There is much to be said for transnational regulation, but we should start it by our own set of principles.

My Lords, the smartphone, the web and ancillaries would have seemed magical just a generation ago. There is no gainsaying their benefits to billions worldwide. It is welcome that machines supplement, if not replace, white-collar jobs, routine legal work, accountancy and even surgery. This clearly needs regulation at a national level.

However, looking ahead, we must keep our minds open, or at least ajar, to transformative advances that may seem exhilarating today but are frightening too. LLMs will surely confront us, writ large with the downsides of existing social media—fake news, photos and videos of unmoderated extremist diatribes and so forth. Social media can spread panic and rumour, and psychic and economic contagion, literally at the speed of light.

Experts such as Geoff Hinton speak of “human extinction”. This may be an exaggeration. My concern is less the science fiction scenarios of a takeover by superintelligence, but rather the risks of breakdowns or sabotage of interconnected networks, electricity grids, GPS, the internet et cetera, which could cause a societal breakdown that cascades globally. Regulation for this is harder. The Atomic Energy Authority works because building an atom bomb requires large-scale conspicuous facilities. It is not so easy to control developers of rogue viruses or, indeed, cyber criminals, and that is the problem we face.

There needs to be a balance to enable innovation to continue but also display a method of doing all we can to prevent such spreads, which could be catastrophic. To quote a well-known proverb, although this is unfamiliar, it is not improbable, and the biggest scare we face in coming decades is this one.

My Lords, I am grateful to the noble Lord, Lord Holmes, for his persistent leadership in this debate. As the Government consider cross-sector AI regulation, I draw your Lordships’ attention to the work I have witnessed as a co-chair of the All-Party Parliamentary Group on Digital Identity and a member of the All-Party Parliamentary Group on Cyber Security and Business Resilience, which frequently fill committee rooms with UK experts and businesses highlighting the opportunities presented by AI and its profound implications for sovereignty, security, democratic accountability and workforce development.

The Government have chosen—perhaps unwisely, in my view—to favour cross-sector regulation rather than introducing a comprehensive AI framework, for which the noble Lord, Lord Holmes, and others have called. Coming as I do from a local government management background, fragmentation was often regarded as a source of inefficiency, inconsistency and risk. Yet AI systems now operate across our healthcare, finance, education and public services through outsourced and disparate technology providers. What assessment have Ministers made of its effectiveness? Has there been a cost-benefit analysis of these proposed, fragmented regulatory approaches?

Other jurisdictions are moving more decisively. The EU has the European Artificial Intelligence Act, with a common framework based on risk and accountability. Singapore has a national AI governance model that combines innovation with oversight, while, as has been said, the United States has been aligning AI development with national security and sovereign capability.

Against this backdrop, what assessment have the Government made of the UK’s position in the Sovereign AI Power Index, which was launched in the UK and which measures national capability across compute, data, talent, energy, research and governance? Do the Government regard strengthening sovereign AI capability as a strategic national priority? Do they consider the stewardship of strategically important data sets under UK sovereign control to be critical to national resilience and security, given the increasing role of global technology companies in managing public infrastructures and sensitive public sector data—including in the NHS and HMRC—in partnership with companies such as Apple, Google and Palantir? They may bring expertise and innovation, but what safeguards exist to ensure that the UK’s most valuable public data assets remain subject to British law and sovereignty, ultimately in the interests of British citizens? I look forward to the Minister’s reply.

My Lords, I declare an interest as an adviser to DLA Piper on AI policy and regulation. We should all thank the noble Lord, Lord Holmes, for his consistent advocacy for regulation, the need for which is clearly shared widely around this Room.

Geoffrey Hinton, the Nobel laureate and godfather of AI, and Yoshua Bengio, the world’s most cited computer scientist, are not alarmists about AI. They are the people who built it, and now, of course, they are our religious leaders. When they call for binding regulation, the Government should listen. Moreover, the Ada Lovelace Institute has found that 89% of the public support an independent AI regulator with enforcement powers, and that 48% reject lighter rules to keep pace with other countries.

This is a manifesto commitment abandoned without explanation. Binding regulation was promised in 2024 and reaffirmed in the King’s Speech thereafter, but it has gone by 2026. The Government say that the existing frameworks suffice. We have the CMA’s conduct requirement for Google but, in other areas, Amazon’s cloud businesses, say, remain unregulated under the Digital Markets Unit after years of investigation. The existing frameworks are not sufficient, and now the competition reform Bill further threatens the independence of the CMA.

On the regulating for growth Bill, the King’s Speech briefing notes make clear that successful sandbox pilots could lead to law being permanently disapplied. This risks becoming a Henry VIII power grab. We await the Bill text, but the stated intention alone should alarm us.

On copyright, 274 commercial licensing agreements between content providers and AI developers already exist. The myth that legal licensing is impossible has always been false. The Government know this, yet even requiring web crawlers to identify themselves has been sidelined. I ask a Minister one question: the Government have the legislative moment, the mandate and their own manifesto; why not bring forward the cross-sector framework that the House, the public and the experts have all called for? The window is still open but, in my view, not for long without huge risks to our society.

I add my thanks to my noble friend Lord Holmes for bringing this debate forward today. It is a complex area, and we would all agree that it is difficult to do it justice with such limited time. I think we also all agree that there are some areas where we definitely need controls on AI: safeguarding against suicidal ideation, imparting dangerous knowledge on biological and other weaponry, and nudification of people, to name just a few areas.

However, we also see that there are dangers to overregulation and tying our hands behind our back. Unfortunately, whether we like it or not, AI is a winner-takes-all world. In the area of defence, as the noble Baroness, Lady Antrobus, mentioned, we need only to see a small technologically advanced Spanish army defeating a much larger Aztec army to learn some of the lessons—we do not want to be in that position as we go forward. Unless we can get every country in the world—and every terrorist group—to sign up to the same set of regulatory rules, we are danger of being on the wrong side of history. I would much rather make sure that we are at the forefront of AI development, rather than at the mercy of others, by tying our hands behind our back with regulatory restrictions.

It is with this knowledge that the Conservative Government took the approach they did, setting up the AI Security Institute to try to safeguard and test models for the reputable producers of AI, so we can make sure they are working in the way we want them to. In combining that with an agile regulatory framework, as my noble friend Lord Camrose previously noted, the priority was to ensure coherence across existing regulations while having a period of non-statutory implementation. AI will touch every aspect of our lives: business, social, health, education, defence and media. I believe it is impossible for one AI regulator to act across all those fields, but much more manageable to make every regulator responsible for their part of AI in that area.

I am afraid, however, that I do not know what this Government’s approach is. Do they intend to continue with the current framework or a more cross-sectoral model? It is vital to get this right to ensure that the UK is at the forefront of AI, so I look forward to hearing from the Minister the Government’s approach.

My Lords, I am pleased to respond to this Question for Short Debate and I am thankful to the noble Lord, Lord Holmes, for initiating this debate, and for the wide variety of contributions made so concisely. It is such an important topic that if I fail to respond, I will of course follow up in writing.

The Government believe that AI has transformative potential for the UK: from scientific innovation and public sector reform, to increasing productivity to drive economic growth. To realise these benefits, we need to make sure that AI is used in a secure and controllable way. Our approach to regulation must enable innovation and protect our citizens from the risks that emerge as AI capabilities develop.

Recognising the point that noble Lords have made, that regulation and growth are not a trade-off, our regulating for growth Bill was announced last month in the King’s Speech. It is an example of how we will make the UK’s regulatory system fit for the future, so that it plays a full role in delivering growth and supporting innovation, including in AI, safely and sustainably. The Bill will create cross-economy sandboxing powers, so that businesses can test cutting-edge new products and technologies safely, prove what works, and then scale up delivery of these changes more quickly. That is how we regulate well and within a controlled environment.

As noble Lords have asked, we do believe that AI is a general-purpose technology with a wide range of applications, which is why the Government believe that most AI systems should be regulated at the point of use. Following the AI Opportunities Action Plan, the Government are committed to working with regulators to boost those capabilities. As a part of this, the Government issued letters to 19 regulators in January 2026, asking them to publish a plan setting out how they will enable safe AI-powered innovation. These regulators cover several high-potential sectors for AI innovation, including life sciences.

In addition, the Regulatory Innovation Office, which was launched in 2024, delivers targeted funding for regulatory experimentation through the regulators’ pioneer fund and the AI capability fund to support pilots, sandboxes and new regulatory pathways. For example, the MHRA is developing an AI-based tool to analyse drug-to-drug interactions in cardiovascular patients, and this will enable MHRA regulators to safely analyse the implications of proposed new drugs quickly and more effectively.

To respond to the question asked by the noble and learned Lord, Lord Thomas, we believe that AI in legal services is a multifaceted area and that a single AI regulator in legal services would risk duplicating existing regulatory functions, creating uncertainty. As I said, we have also written to the Legal Services Board asking for its plan on how to regulate AI safely.

As my noble friend Lady Antrobus mentioned, the issues of defence are complex. AI in defence is moving very fast. It is becoming a defining feature of modern warfare and, as she mentioned, a critical enabler of defence capability. We have an updated strategic approach to AI in defence, which reflects the more operational delivery-focused model. We are prioritising AI-enabled war-fighting advantage and enterprise transformation, strengthening governance and accountability, and ensuring that ambition is backed by the data, compute, skills and partnerships needed to scale at pace while taking a disciplined approach to frontier AI and maintaining robust standards. We also remain firmly committed to context-appropriate human involvement, which centres on humans as the accountable actors in the use of force. The role of any human intervention is therefore to ensure that responsibility and accountability are also clearly retained by people and not machines, to one of the important points my noble friend raised.

The noble Baroness, Lady Uddin, made a very important point about the fact that regulation does not happen without the context of the other measures we are taking to support the development of UK-based AI. We have launched our sovereign AI fund, supporting the development of AI capability here, as well as our support for the AI growth zones, as noble Lords will know. In addition, we are investing in the skills of people in this country. Effective regulation will only go hand in hand with a workforce, regulators and everybody being skilled to understand the risks and the judgments that have been taken day to day—including in media literacy, being able to take that discerning view about what people are seeing and consuming as part of media.

On the important points made about the fact that this is a global set of developments, and on the important role the UK has taken in the past and continues to take in shaping the passage of key international AI initiatives, we have indeed led on initiatives such as the Global Dialogue on AI Governance and the Independent International Scientific Panel on AI at the UN, and the Framework Convention on Artificial Intelligence in the Council of Europe. These are really important initiatives. Earlier this week I was at the OECD, where we as the UK were supporting the dissemination of principles for AI policy-making. It is very important that all countries apply regulation and policy effectively, as noble Lords have said. This is a globally developing set of initiatives.

Many noble Lords, including the noble Lords, Lord Harper, Lord Tarassenko and Lord Markham—and I heartily agree with them—mentioned the AI Security Institute. This is indeed an institute that the Government are proud of. It is world-leading and a centre of expertise, and has been analysing AI systems for two and half years. It is in an extremely unusual position globally, in that it has close collaboration with AI labs and has tested over 30 models to understand their potentially harmful capabilities. Leading industry players, including Anthropic and OpenAI, have made changes to strengthen AI model safeguards based on the institute’s findings.

This foundational research—to discover methods for building AI systems that are beneficial, reliable and aligned with human values—is essential, and these findings are shared with the Home Office, the NCSC and other national security organisations, enabling the UK to stay one step ahead of the risks brought by AI capabilities. We are committed to giving the institute the funding it needs through the spending review.

Noble Lords have highlighted the speed of development of AI models, and earlier this year Anthropic announced that Mythos represented a significant step-up in AI cyber capabilities. This is being monitored carefully by AISI and the NCSC, and they have published their findings on that. We have given further guidance to businesses and regulators on the measures that need to be taken to prevent and mitigate the risks of these further developments.

The noble Lord, Lord Taylor, mentioned the Cyber Security and Resilience (Network and Information Systems) Bill, which is moving to Report in the other place this month. It is designed to protect the services that the public rely on, and those regulations take an all-hazards risk-based approach that requires organisations to manage cyber security, physical security and broader operational risks, instead of specifying particular risks of technologies, in order to stay up to date and take into account further developments in technologies of the future.

I fear that I am not going to do justice to the topic of copyright.

Just before the Minister sits down—we have seven minutes left before the debate runs out—can I ask her a question? My noble friend Lord Holmes drew attention to the Government’s original commitment to legislate for a cross-sector approach, and he referenced how the Government have now dropped that commitment. The only bit of legislation that the Government are doing is, as the Minister said, in the regulation for growth Bill, but I have not heard the Government explain why they have changed their approach from the one they originally set out. Can the Minister, in the seven minutes we have left, set out for noble Lords the reasons why the Government have done that? That would be helpful.

Our view is that AI is a general-purpose technology with a wide range of applications, and also that regulators understand well the sectors they are regulating. They understand the risks that are present and understand, as many noble Lords have raised in the debate, the specific applications that AI is bringing, the potential risks to consumers and the nature of the competitive landscape—the balance between competition, incumbents and new entrants. That is the reason we are taking the approach to regulate at the point of use and using those regulatory frameworks.

We also already regulate AI in the UK—for example, in the Online Safety Act or under the GDPR—and, in some areas, across sectors. There are some areas in which legislation applies to the application of AI technologies.

On that question of going domain-specific, what occurs in situations where there is a domain with no competent regulator? Similarly, how do the Government assure clarity, consistency and coherence of approach? As a citizen, you may come across AI—as noble Lords have rightly identified—in defence, education, health, tax and benefits. If you do not some sense of a guiding mind or horizontal direction, how can you possibly deliver consistency and clarity for people wherever they may come across AI?

In all regulation, there is obviously a balance between consistency and context-specific, appropriate regulation. It is not always the case that consistency is the most appropriate or first-order principle. It may be that, as we have been discussing, there are many issues: a focus on growth, a focus on consumer protection or, for the energy markets in particular, a motivation towards decarbonisation. That is why the regulator, for whichever market we are talking about, is very well placed to look at how their objectives, as set out in their statutory duties, are best applied in the context of this new technology, which provides different functionalities and the opportunity for new innovation.

That is one reason we have also given these sandboxing powers. We realise that the current set of statutory frameworks was set up assuming that humans would always, for example, be in vehicles or crew vessels. We may need to adapt that in order to take account of the potential new innovations that AI brings, while doing so in a safe and secure way.