End of business adjournment debate: National AI Strategy and UNESCO AI Ethics Framework - May 22

Darren Jones (Bristol North West) (Lab)

I am grateful, Mr Deputy Speaker, that this Adjournment debate on the regulation of artificial intelligence has been granted. I declare my interest as set out in the Register of Members’ Financial Interests.

Britain is at a turning point. Having left the European Union, irrespective of what people thought about that decision, we have decided to go it alone. This new chapter in the long history of our great nation is starting to unfold, and we have a number of possible destinations ahead. We stand here today as a country with great challenges and an identity crisis: what is modern Britain to become? Our economy is, at best, sluggish; at worst, it is in decline. Our public services are unaffordable, inefficient and not delivering the quality of service the public should expect. People see and feel those issues right across the country: in their pay packets, in the unfilled vacancies at work, and in their local schools, GP surgeries, dentists, hospitals and high streets. All of this is taking place in a quickly changing world in which Britain is losing influence and control, and for hostile actors who wish Britain—or the west more broadly—harm, those ruptures in the social contract present an opportunity to exploit.

Having left the European Union, I see two destinations ahead of us: we can either keep doing what we are doing, or modernise our country. If we take the route to continuity, in my view we will continue to decline. There will be fewer people in work, earning less than they should be and paying less tax as a consequence. There will be fewer businesses investing, meaning lower profits and, again, lower taxes. Income will decline for the Treasury, but with no desire to increase the national debt for day-to-day spending, that will force us to take some very difficult decisions. It will be a world in which Britain is shaped by the world, instead of our shaping it in our interests.

Alternatively, we can decide to take the route to modernity, where workers co-create technology solutions at work to help them be more productive, with higher pay as a consequence; where businesses invest in automation and innovation, driving profits and tax payments to the Treasury; where the Government take seriously the need for reform and modernisation of the public sector, using technology to individualise and improve public services while reducing the cost of those services; and where we equip workers and public servants with the skills and training to seize the opportunities of that new economy. It will be a modern, innovative Britain with a modern, highly effective public sector, providing leadership in the world by leveraging our strengths and our ability to convene and influence our partners.

I paint those two pictures—those two destinations: continuity or modernity—for a reason. The former, the route to continuity, fails to seize the opportunities that technological reforms present us with, but the latter, the route to modernity, is built on the foundations of that new technological revolution.

This debate this evening is about artificial intelligence. To be clear, that is computers and servers, not robots. Artificial intelligence means, according to Google,

“computers and machines that can reason, learn, and act in such a way that would normally require human intelligence or that involves data whose scale exceeds what humans can analyse.”

These AI machines can be categorised in four different ways. First, reactive machines have a limited application based on pre-programmed rules. These machines do not use memory or learn themselves. IBM’s Deep Blue machine, which beat Garry Kasparov at chess in 1997, is an example. Secondly, limited memory machines use memory to learn over time by being trained using what is known as a neural network, which is a system of artificial neurons based on the human brain. These AI machines are the ones we are used to using today. Thirdly, theory of mind machines can emulate the human mind and take decisions, recognising and remembering emotions and reacting in social situations like a human would. Some argue that these machines do not yet exist, but others argue that AI such as ChatGPT, which can interact with a human in a humanlike way, shows that we are on the cusp of a theory of mind machine existing. Fourthly, self-aware machines are machines that are aware of their own existence and have the same or better capabilities than those of a human. Thankfully, as far as I am aware, those machines do not exist today.

That all might be interesting for someone who is into tech, but why am I putting it on the public record today? I am doing so because there are a number of risks that we as a Parliament and the Government must better understand, anticipate and mitigate. These are the perils on our journey to continuity or modernity. Basic artificial intelligence, which helps us to find things on the internet or to book a restaurant, is not very interesting. The risk is low. More advanced artificial intelligence, which can perform the same tasks as a junior solicitor, a journalist or a student who is supposed to complete their homework or exam without the assistance of AI, presents a problem. We already see the problems faced by workers who have technology thrust upon them, instead of being consulted about its use. The consequences are real today and carry medium risks—they are disruptive.

Then we have the national security or human rights-level risks, such as live facial recognition technologies that inaccurately identify someone as a criminal, or a large language model that can help a terrorist understand how to build a bomb or create a novel cyber-security risk, or systems that can generate deepfake videos, photos or audio of politicians saying or doing things that are not true to interfere with elections or to create fake hostage recordings of someone’s children.

Jim Shannon (Strangford) (DUP)

I commend the hon. Gentleman on bringing this debate forward. It is a very deep subject for the Adjournment debate, but it is one that I believe is important. Ethics must be accounted for to ensure that any industries using AI are kept safe. One issue that could become increasingly prominent is the risk of cyber-threats, which he referred to, and hacking, which not even humans can sometimes prevent. Does he agree that it is crucial that our Government and our Minister undertake discussions with UNESCO, for example, to ensure that any artificial intelligence that is used within UK industry is assessed, so as to deal with the unwanted harms as well as the vulnerabilities to attack to ensure that AI actors are qualified to deal with such exposure to cyber-attacks? In other words, the Government must be over this issue in its entirety.

Darren Jones

The hon. Member is of course right. In the first part of his intervention, he alluded to the risk I have just been referring to, where machines can automatically create, for example, novel cyber-risks in a way that the humans who created those systems might not fully understand and that are accessible to a wider range of actors. That is a high risk that is either increasingly real today or is active and available to those who wish to do us harm.

The question, therefore, is what should we in Parliament do about it? Of course, we want Britain to continue to be one of the best places in the world to research and innovate, and to start up and scale up a tech business. We should also want to transform our public services and businesses using that technology, but we must—absolutely must—make sure that we create the conditions for this to be achieved in a safe, ethical and just way, and we must reassure ourselves that we have created those conditions before any of these high-risk outcomes take place, not in the aftermath of a tragedy or scandal.

That is why I have been so pleased to work with UNESCO, as the hon. Gentleman mentioned, and assistant director general Gabriela Ramos over the past few years, on the UNESCO AI ethics framework. This framework, the first global standard on AI ethics, was adopted by all 193 member states of the United Nations in 2021, including the United Kingdom. Its basis in human rights, actionable policies, readiness assessment methodology and ethical impact assessments provides the basis for the safe and ethical adoption of AI across countries. I therefore ask the Minister, in summing up, to update the House on how the Government are implementing their commitments from the 2021 signing of the AI ethics framework.

As crucial as the UNESCO AI ethics framework is, in my view the speed of innovation requires two more things from Government: first, enhanced intergovernmental co-ordination, and secondly, innovation in how we in this House pass laws to keep up with the speed of innovation. I will take each in turn.

First, on enhanced intergovernmental co-ordination, I wrote to the Government at the end of April calling on Ministers to play more of a convening role on the safe and secure testing of the most advanced AI, primarily with Canada, the United States and—in so far as it can be achieved—China, because those countries, alongside our own, are where the most cutting-edge companies are innovating in this space. I was therefore pleased to see in the Hiroshima communiqué from last week’s G7 a commitment to

“identify potential gaps and fragmentation in global technology governance”.

As a parliamentary lead at the OECD global parliamentary network on AI, I also welcome the request that the OECD and the Global Partnership on Artificial Intelligence establish the Hiroshima AI process, specifically in respect of generative AI, by the end of this year.

I question, however, whether these existing fora can build the physical or digital intergovernmental facilities required for the safe and secure testing of advanced AI that some have called for, and whether such processes will adequately supervise or have oversight of what is taking place in start-ups or within multinational technology companies. I therefore ask the Minister to address these issues and to provide further detail about the Hiroshima AI process and Britain’s contribution to the OECD and GPAI, which I understand has not been as good as it should have been in recent years.

I also welcome the engagement of the United Nations’ tech envoy on this issue and look forward to meeting him at the AI for Good summit in Geneva in a few weeks’ time. In advance of that, if the Minister is able to give it, I would welcome his assessment of how the British Government and our diplomats at the UN are engaging with the Office of the Secretary-General’s Envoy on Technology, and perhaps of how they wish to change that in the future.

Secondly, I want to address the domestic situation here in the UK following the recent publication of the UK’s AI strategy. I completely agree with the Government that we do not want to regulate to the extent where the UK is no longer a destination of choice for businesses to research and innovate, and to start up and scale up their business. An innovation-led approach is the right approach. I also agree that, where we do regulate, that regulation must be flexible and nimble to at least try to keep up with the pace of innovation. We only have to look at the Online Safety Bill to learn how slow we can be in this place at legislating, and to see that by the time we do, the world has already moved on.

Where I disagree is that, as I understand it, Ministers have decided that an innovation-led approach to regulation means that no new legislation is required. Instead, existing regulators—some with the capacity and expertise required, but most without—must publish guidance. That approach feels incomplete to me. The European Union has taken a risk-based approach to regulation, which is similar to the way I described high, medium and low-risk applications earlier. However, we have decided that no further legislative work is required while, as I pointed out on Second Reading of the Data Protection and Digital Information (No. 2) Bill, deregulating in other areas with consequences for the application of consumer and privacy law as it relates to AI. Surely, we in this House can find a way to innovate in order to draft legislation, ensure effective oversight and build flexibility for regulatory enforcement in a better way than we currently do. The current approach is not fit for purpose, and I ask the Minister to confirm whether the agreement at Hiroshima last week changes that position.

Lastly, I have raised my concerns with the Department and the House before about the risk of deepfake videos, photo and audio to our democratic processes. It is a clear and obvious risk, not just in the UK but in the US and the European Union, which also have elections next year. We have all seen the fake picture of the Pope wearing a white puffer jacket, created by artificial intelligence. It was an image that I saw so quickly whilst scrolling on Twitter that I thought it was real until I stopped to think about it.

Automated political campaign videos, fake images of politicians being arrested, deepfake videos of politicians giving speeches that never happened, and fake audio recordings are already available. While they may not all be of perfect quality just yet, we know how the public respond to breaking news cycles on social media. Many of us look at the headlines or the fake images over a split second, register that something has happened, and most of the time assume it to be true. That could have wide-ranging implications for the integrity of our democratic processes. I am awaiting a letter from the Secretary of State, but I am grateful for the response to my written parliamentary question today. I invite the Minister to say more on that issue now, should he be able to do so.

I am conscious that I have covered a wide range of issues, but I hope that illustrates the many and varied questions associated with the regulation of artificial intelligence, from the mundane to the disruptive to the risk to national security. I welcome the work being done by the Chair of the Science, Innovation and Technology Committee on this issue, and I know that other Committees are also considering looking at some of these questions. These issues warrant active and deep consideration in this Parliament, and Britain can provide global leadership in that space. Only today, OpenAI, the creator of ChatGPT, called for a new intergovernmental organisation to have oversight of high-risk AI developments. Would it not be great if that organisation was based in Britain?

If we get this right, we can take the path to modernity and create a modern Britain that delivers for the British people, is equipped for the future, and helps shape the world in our interests. If we get it wrong, or if we pick the path to continuity, Britain will suffer further decline and become even less in control of its future. Mr Deputy Speaker, I pick the path to modernity.

9.22pm
The Parliamentary Under-Secretary of State for Science, Innovation and Technology (Paul Scully)

I congratulate the hon. Member for Bristol North West (Darren Jones) on securing this excellent debate and on his excellent opening speech. The issue ahead of us is an international issue, and as he said, the UK is at the forefront of AI development, with our history and with the Turing and Lovelace institutions around the country. We have amazing AI clusters, and it is right that we should be at the forefront of the solutions he talked about. It will not have escaped many of us with a long-standing interest in AI that this is a really important time for the technology’s development. Of equal note is the focus that the Government are giving to ensuring that we seize the opportunities of AI while tackling the risks that have been highlighted, along with our commitment to iterating and adapting our approach as the technology continues to develop.

I welcome the opportunity to speak about how we are delivering on the commitments of the national AI strategy, including shaping the international governance of AI through active engagement in key multilateral fora such as UNESCO. I believe we are well placed to become a global AI superpower by delivering on the foundations laid down in the national AI strategy and its three pillars: investing in and planning for the long term needs of the AI ecosystem; supporting the transition to an AI-enabled economy, capturing the benefits of innovation in the UK and ensuring that AI benefits all sectors and regions; and ensuring that the UK gets the national and international governance of AI technologies right to encourage innovation and investment and to protect the public and our fundamental values.

The Government recognise that AI has the potential to transform all areas of life, from making more medical breakthroughs possible to powering the next generation of tech such as driverless cars. In 2021 we published our national AI strategy—a 10-year vision to make the UK an AI superpower. Since 2014, we have invested over £2.5 billion in AI, including almost £600 million towards the near £1 billion 2018 AI sector deal, which kick-started the growth of the already well-established AI landscape in the UK; £250 million to develop the NHSX AI lab to accelerate the safe adoption of AI in health and care; £250 million for the Centre for Connected and Autonomous Vehicles to develop the future of mobility in the UK; investment in the Alan Turing Institute, with over £46 million to support Turing AI fellowships to develop the next generation of top AI talent; and over £372 million of investment in UK AI companies through the British Business Bank.

The AI strategy also emphasises the need to invest in skills and diversity to broaden the AI workforce. Our £30 million AI and data science conversion course and scholarship programme was set up to address the lack of diversity and supply of talent in the UK AI labour market—that is diversity not as in a tick-box exercise, as some might be, but diversity of thinking to ensure that AI products, services and development have the broader thinking that the hon. Member rightly talked about.

Alongside skills, the Government recognise the need for long-term investment in computing. In March, we announced £900 million for an exascale supercomputer and AI research resource. Building on that, last month we announced £100 million in initial start-up funding for a foundation model taskforce to invest in the AI stack to build foundation model capability, ensure capabilities for key use cases and ensure UK leadership in the safety and reliability of foundation models.

We have seen huge leaps forward in our delivery on the governance pillar of the national AI strategy. In March, we published a White Paper setting out the UK’s context-based, proportionate and adaptable approach to AI regulation, representing a world-leading step forward in this policy space. The White Paper outlines five clear outcome-focused principles that regulators should consider to facilitate the safe and innovative use of AI in the industries that they monitor. Crucially, the principles provide clarity to businesses by articulating what we want responsible AI to look like.

That is not all. In October 2022, we launched the AI standards hub to increase the UK’s contribution to the development of global AI technical standards. Through the hub, we are working with international initiatives such as the OECD’s catalogue of tools and metrics for trustworthy AI to increase global awareness of technical standards as critical tools to advance the worldwide development and adoption of responsible AI.

On that note, I turn my focus squarely to international engagement on AI, which is a key priority for the Government. As a world leader in AI, we play an important role in shaping the international development and governance of AI. We promote our interests in bilateral relationships with key partners such as the US and Japan and in multilateral fora such as the Council of Europe, the Global Partnership on Artificial Intelligence, UNESCO, the OECD, the G7, the International Organisation for Standardisation and International Electrochemical Commission.

With the US, we held the inaugural meeting of the comprehensive dialogue on technology and data in January. A key deliverable for 2023 is to strengthen the UK-US collaboration on AI technical standards development and tools for trustworthy AI, including through joint research and information sharing, and support for commercial co-operation. We had previously signed in September 2020 a US-UK declaration on co-operation in AI research and development, representing a shared vision for driving technological breakthroughs in AI. With Japan, as the hon. Member rightly said, we agreed the Hiroshima accord only recently, on 18 May. It is a landmark new global strategic partnership, signifying our intent to work together to maintain strategic advantage in emerging technologies such as AI. The accord builds on the UK-Japan digital partnership that I launched in December 2022, which established a framework for deeper UK-Japan collaboration across digital infrastructure and technologies, data, digital regulation and digital transformation.

We have also been working closely with Japan as part of its G7 presidency this year. At the end of April, I attended the G7 digital ministerial meeting in Japan, where I signed the G7 digital ministerial declaration alongside my counterparts. That declaration emphasises the importance of responsible AI and global AI governance. It endorses an action plan for promoting global interoperability between tools for trustworthy AI and for co-operating on upcoming AI opportunities and challenges.

At the Council of Europe, we are working closely with like-minded nations on the proposed convention on AI—a first-of-its-kind legal agreement to help protect human rights, democracy and the rule of law. At the OECD, we are an active member of the working party on AI governance, which supports the implementation of the OECD’s AI principles. It enables the exchange of experience from best practice to advance the responsible stewardship of AI. At the global partnership, we are a key contributor and founding member. At the 2022 GPAI ministerial summit in Japan, we announced £1.2 million of funding to develop a net zero data space for AI applications, which is in addition to a previous £1 million investment to advance GPAI research on data justice, collaborating with our world-leading Alan Turing Institute and 12 pilot partners in low and medium-income countries.

We are also leading the development of global AI technical standards in standards development organisations such as the International Organisation for Standardisation and the International Electrotechnical Commission, and we are leading the development of AI assurance techniques as additional tools for trustworthy AI. Crucially, these techniques help to measure, evaluate and communicate the trustworthiness of AI systems across the development and deployment life cycle, to enable organisations to determine whether AI technologies are aligned with regulatory requirements.

We are also aware of the increasing prominence of AI in discussions held across other UN fora, including the Internet Governance Forum and the International Telecommunication Union, and through the Global Digital Compact’s focus on AI. The Government welcome the opportunity that the compact provides for the multi-stakeholder community to set out an ambitious shared agenda, chart a path for concrete action towards delivering it, and promote the sharing of best practice, evidence and learning.

Let me turn my attention to UNESCO. The UK was actively involved in the development of its recommendation on the ethics of AI, and UK organisations such as the Alan Turing Institute have supported the development of implementation tools. As we have heard, we, along with all 192 other UNESCO member states, adopted the recommendations in November 2021, demonstrating our commitment to developing a globally compatible system of responsible and ethical AI governance.

Our work aligns with the values of UNESCO’s recommendation. For example, through our work at the Council of Europe negotiations, we are helping to respect, protect and promote human rights, fundamental freedoms and human dignity. In doing so through close collaboration with our international partners, we aim to ensure that our citizens can live in peaceful, just and interconnected societies. Through our AI and data science conversion course and scholarship programme, we are ensuring diversity and inclusiveness by addressing these issues in the UK AI labour market. Finally, as one small example of the wider work we are delivering, through our net zero data space for AI applications, funded through GPAI, we are delivering on our net zero policy objectives, ensuring a flourishing environment and ecosystem.

In summary, we have taken great strides in our delivery of the national AI strategy under all three pillars: investing in and planning for the long-term needs of the AI ecosystem; supporting the transition to an AI-enabled economy; and ensuring that the UK gets the national and international governance of AI technologies right. It goes without saying that the opportunities afforded by AI are quite staggering. Indeed, as a result of AI technologies, UK productivity could rise by up to a third across sectors, and UK GDP could be 10.3% higher in 2030 as a result of AI—the equivalent of an additional £232 billion.

But the hon. Gentleman is also absolutely right to look at the risks and talk about the dangers. We have to do this on an international basis. The AI White Paper was the first of its kind, although I would urge him to exercise caution when he says that we do not feel that we need legislation. At the moment, we are building on the layers of existing regulation, but the White Paper outlines the five principles, and we are looking at the regulatory sandboxes to test regulation with scientists, the sector and the academics involved, so that we can co-create the solutions that will be required. But we clearly have to do this at pace, because it was only a few months ago that we first heard of ChatGPT, and we now have prompt engineers, a new, relatively well paid occupation that until recently no one had ever heard of.

As a world leader in AI, it is imperative that we continue to actively engage bilaterally and in multilateral fora such as UNESCO, but also in the OECD, the GPAI and others, to shape the international AI governance landscape. Governing it effectively will ensure that we achieve the right balance between responding to risks and maximising the opportunities afforded by this transformative technology.

Question put and agreed to.