25
Examination of witnesses
Rt Hon Michelle Donelan MP and Emran Mian.
Q1 The Chair: This is the Communications and Digital Committee. This is a one-off hearing with the Secretary of State for Science, Innovation and Technology, Michelle Donelan, accompanied by Emran Mian, the director-general for digital technologies and telecoms at DSIT. We are very grateful and very pleased to have you here, Secretary of State. This is the first time that we have been able to see you before us since you took up the position at DSIT.
During that time, we have undertaken two major inquiries: first, into digital exclusion, and, more latterly, into large language models and generative AI. We will raise questions covering both of those inquiries. We will come on to digital exclusion later. We will start with large language models and generative AI.
We know, of course, that today you published your White Paper response, and we are very pleased that it has been possible for you to do so. No doubt, you will draw on that in your responses to some of the questions that we ask. We have not had too much time to look at it and study it in detail, but I see that there is quite a lot of common ground and some things that are supportive of similar findings that we made in our report. In light of that, what made the department put out a statement on Friday rejecting the findings of this committee’s report?
Michelle Donelan MP: There is a lot of commonality in the report that you issued. We have addressed different areas of the themes that you covered. However, in entirety, the two do not align perfectly. What I took from your report was very much that there needs to be more emphasis on the risks and gathering the evidence base, and that is certainly something that we are pursuing. At the heart of our response to our White Paper is that we are doubling down on what we set out last year, continuing down a road of proportionality, and making sure that we support our regulators not just with words but with actions.
We announced a £10 million pot to help to support them, including with skills for instance, and we are making sure that the current system of regulators and regulation is working to the same principles around safety, proportionality and fairness. We are examining whether there are any potential gaps and what more we need to do to fill them and whether that includes legislation or additional support in the form of manpower or what have you, and beefing up our central risk function. There is commonality, but the two do not identically link.
The Chair: No, of course. We will come on to the specifics of risks and the approach that you are taking on regulation and how you are supporting the regulators to equip themselves better to do what is required of them. In fact, on some of the approaches to risks there was agreement. One of our main findings was that within government there had perhaps been too narrow a focus on the high end of safety issues and not enough on opportunities. We made some specific points on risks, which we will come to later, on ensuring that the near-term risks are prioritised and on what more might need to be done on catastrophic risk. It was about the balance between safety and innovation where I thought there was a difference.
Michelle Donelan MP: On that I disagree, because we have been prioritising innovation. They are two sides of the same coin. If you look at any speech that I or one of my Ministers has ever made, or any comments that we have ever given, we always talk about the huge opportunities, because that is the reason why we need to grip the risks—so that it is safe to access the opportunities and to ensure that we take people with us and we get public trust. For me, one of the biggest risks that AI presents is that we turn people away from the potential and they become too scared to adopt it or that we see slow adoption in our businesses and our public services, and then we will not get those fantastic opportunities.
At the heart of our entire approach is trying to bolster the industry itself to develop innovation so that we not only have the ideas created but we ensure that the jobs are here in the UK, and that we benefit as an economy. This is not just what we are saying as a Government, it is what many people in the sector tell us; in many of the different quotes that we issued today along with our White Paper, the response is that this is a pro-innovation approach when it comes to AI that is remarkably different from the tack that other nations and nation groups such as the EU have taken.
I reject the premise that at one time we are picking one over the other, because it is about both at the same time. We will continue to try to lean into both in our language and in our actions.
Q2 The Chair: I think you are right. We were clear too that it is about getting the balance right. It is not a zero-sum game where there is a choice between one or the other. The evidence that we took through the course of the inquiry seemed to signal a shift by the Government to prioritising safety over innovation, and that was notable from the way in which things had developed from the point at which you published the White Paper in March last year, with the focus on moving much more exclusively to safety, as your work and priorities were talked about and put forward from the Government.
Michelle Donelan MP: I do not see it that way. As I said, the whole point of gripping the potential risks and the risks is so that we can seize the opportunities. When we did the summit, we did not just talk about the risks; we made sure that there were big agenda items on AI for good and the opportunities that that presented, including for the developing world. We had a whole section on that. We looked at how we can utilise AI more in our public services, including our education system. We made two big announcements about money going into AI funded by the Government for education to support our teachers. We made £100 million available for an AI accelerator to support our healthcare missions.
While the media coverage might sometimes lurch towards a more catastrophic element of this topic—because, let us face it, sometimes that can produce clickbait headlines—the Government’s own language and approach have always been mindful of the fact that this technology is two sides of the same coin and we do not want to turn people away from the technology and frighten them. We actually want to be saying, and sending out a clear message, that there are very real risks. We published the risk document to be fully transparent with the public. We are the first country in the world to do that, and I think we still are to date. We need to grip those risks so that we can seize the incredible opportunities. We have given so many examples of the amazing opportunities and talked about what it is already doing in our NHS now, detecting breast cancer and being used in over 90% of stroke units, and then looking ahead to what the opportunities could be.
We must note that, from the publication of the White Paper to the timing of the summit, the technology was going great guns. This is emerging quicker than any technology, I would argue, that we have ever seen in history. Therefore, the potential risks and the knowledge base in those potential risks is emerging as well. One of the key things that we said in the White Paper, and we still say, is that we need to maintain an agile approach. That will mean that our methods and tack will change over time, based on the evidence and the knowledge that we have, given how the technology is emerging. What we do not want to do, which is one of the reasons why we are not lurching towards legislation, is bind our hands with something that is out of date by the time that it becomes a law. We want an agile approach. The rhetoric and the language will sometimes change, just as the terminology changes; this is an advancing field of not only science but academia.
The Chair: Certainly, during the course of last year, there was quite a lot of—. I will try not to say the word.
Michelle Donelan MP: I know what you mean.
The Chair: There was real existential robot apocalypse—that is the word I am looking for—talk coming from certain parts of the tech sector, and that was very much the backdrop. It seemed that at the point when you published the White Paper, which referred to providing £100 million for a foundation model taskforce and accelerating the UK’s capability in artificial intelligence and ensuring sovereign capabilities, it was something which was very much front-foot and seizing the opportunity, but by the time that we got through the summer and into the autumn that seemed to have dropped down the agenda and it was much more about safety.
Michelle Donelan MP: We changed our approach slightly in terms of the machinery of government but not in terms of the importance we were placing on the opportunities. We have as a Government set up a new unit, an AI incubator, run from the Cabinet Office and No. 10. That is focusing on how we can deploy AI in our public services effectively, efficiently and quicker, and they have been onboarding people to that. That is in addition to some of the work that we have been doing.
We have thought about where the best place is in government to be doing certain bits and how much focus we can put on each of those, but we definitely have not downgraded how important the opportunities or the innovation are. I would not want anybody to be left with that impression. We will continue to emphasise regularly why we are doing this. The whole point of gripping the risks is so that we can seize the incredible opportunities to enable people literally to live longer, healthier and easier lives.
The Chair: Is the national AI strategy that came out in 2021 still relevant?
Michelle Donelan MP: Things have progressed considerably since 2021. I cannot remember the exact month, but it is three years this year. If we look back at large language models, four years ago they could barely string a coherent sentence together, and now they can pass the Bar and medical exams, so things are marching on dramatically.
If your follow-up question is whether we are going to produce another strategy, I am somebody who is very focused on delivery as opposed to constant strategies. We have done a considerable amount of work on being transparent with the public as to what we are trying to achieve, and we should definitely keep doing that given the magnitude of the topic, but I would not subscribe to another strategy per se in this area.
The Chair: No, I was not going to ask you whether you were going to do a new strategy.
Michelle Donelan MP: That is normally the question I get from everybody on anything.
The Chair: We may come on to that later. I asked about that, because it has been pointed to as something that had a more positive vision for the technology, and the fact that it is not featuring in the White Paper, or perhaps even in the response to the White Paper that you published today, is another thing that some people point to as evidence that the Government have shifted away. It comes back to balance all the time. Safety is the primary priority, whereas before it was about opportunity and innovation, and the risks were there to make sure that it was possible to maximise the opportunity.
Michelle Donelan MP: I still disagree with that statement, because I believe that we are prioritising safety to access the opportunities and that they are two sides of the same coin, and we have not picked one over the other. It is right that we are leading on the safety agenda, and that we are leading the world on that, but I would argue that we are also a leader when it comes to innovation in AI. That is the reason why OpenAI and Anthropic have chosen to base their international offices here. It is why Cohere has chosen to locate its CEO here. I was at its Canada office last week and at its UK office this week. We are seen as a hub of innovation. We are seen to be supporting industry in that regard.
We are spending a lot of taxpayer money, as well as time, on working out how we can ensure that the benefits are adopted in our public services and in our businesses. Next week we have the first meeting of the AI opportunity forum, which is a collection of different businesses, and we will be discussing and learning lessons from previous technological innovations as to how we can ensure that we inspire confidence and remove some of the barriers to businesses adopting AI quickly and effectively.
Q3 Lord Hall of Birkenhead: Thank you for being here today. You mentioned some of the things you are doing to boost opportunities around large language models. Could you reflect on what you think are the most effective things that the Government are doing to support those opportunities? We kept coming back to a particular thing in the work that we have been doing, and that is the scale of compute necessary now and in the future, and what the UK needs to do to ensure that we have the compute capacity to enable researchers, universities, businesses and so on to keep pace with our international competitors. The Government have put money into it. Microsoft has put even more money into it. It is almost like an arms race. What is our perspective on that?
Michelle Donelan MP: On your first point, when you said opportunities, do you mean opportunities for businesses or for people?
Lord Hall of Birkenhead: Yes.
Michelle Donelan MP: Businesses?
Lord Hall of Birkenhead: For businesses and for universities.
Michelle Donelan MP: On the compute point, you are quite right: compute is an incredibly important part of the AI triad, as it is referred to. I have tried to carve out a three-point strategy around compute. The first point is about the Government investing in compute. We have invested over £1.5 billion, and that includes the AI research resource in Bristol and our exascale plan that will be based in Edinburgh. We will continue to add to that. However, we have to remember the sheer volume of expenditure per year of companies such as Meta in relation to compute. It is not feasible for any Government to get into an arms race to try to match-fund that, so the strategy needs more pillars.
The second area is about trying to attract inward investment from private companies, and we have been successful so far to date: £1 billion was invested by Google and £2.5 billion by Microsoft.
The third part is about how we can pursue partnerships with allied countries in this space. Last week, I co-signed an MoU with Canada on compute, and that will begin a partnership on compute collaboration. We are just fleshing it out and working out how it could look in tangible real terms, and I want that to become almost the blueprint that we can take to other countries, to forge compute collaborations with allied countries, bearing in mind national security considerations. That is our three-tier strategy.
The other point to note when it comes to compute is that one of the things we need to be mindful of is future-proofing our approach to AI and always having foresight and looking ahead to where the industry could develop. While compute is absolutely essential now, it could be that algorithms become more important in the future, which would open up other risks because it would mean that it will be much more readily available for people to access and the barriers to entry of the industry will be reduced, which may have not only a positive effect in equality and democratising it but the risk of bad actors getting a stake.
The other issue is distributed training and, in layman’s terms, instead of just needing one large source of compute whether you could access multiple sources. We have to think not just about the situation that we are dealing with now but about the future, and that is what we are trying to do via the institute and our teams in the department where we have onboarded many experts not just from industry but from civil society.
On the other part of your question, about opportunities for businesses, we have set up the forum. A key thing is the skills base, both from a general public adoption point and from a business adoption point. When I speak to businesses, there is a degree of hesitancy and fear of the unknown, and concern about not wasting money. You will see my department particularly focus on that this year. There is no point our gripping all these risks and talking about all the opportunities unless we actually deliver them in the UK and people actively feel the tangible results. That will mean that we need to work very closely with the Cabinet Office on public services and that we support our businesses, not just large businesses but small businesses too, and we are working very closely with DBT on that.
The broader point about general public adoption will require skills, and we are doing a great deal of work on skills, which I am sure we will come to later. I do not want to spend the entire session talking to this one question; I can feel myself carrying on. We will have spent over £3 billion this Parliament.
The Chair: I am so sorry to interrupt. It was not because you were talking about skills, although we will come on to that. Apparently, the audio feed is not working. Anyone outside this room is not hearing us. I will give our technical colleagues a moment or two to see if they can fix it before we move on to Lord Hall’s next question.
The Committee suspended and then returned.
The Chair: I am sorry for that brief interruption, although anybody watching us online who still cannot hear us is probably none the wiser as to what I am saying. To reassure people—we will make sure this message gets out via Twitter and so on—the on-demand version will be audible and it will be available ASAP once the session has finished. Lord Hall, back to you.
Lord Hall of Birkenhead: That was a really interesting answer, Secretary of State, thank you, particularly on Canada and the notion that nations might come together to get the amount of compute power that we need. There is one little thing on the back of it, though. The £80 million to launch nine research hubs is a lot of money to do something that is clearly quite important, but how do you make a judgment about what matches the need? The need is exponential, in a way. Could you explain that?
Michelle Donelan MP: That was done by UKRI independently. We did not get into the minutiae of how much money is going to Liverpool over London and which universities are getting what for what projects. UKRI has looked at the particular research initiatives and whether they have enough legs to be able to provide that additional funding. You are quite right; you could spend infinite amounts on these types of areas, given their importance now and in the future.
The fact that we are spending so much as a Government highlights the fact that we appreciate how important it is to access the opportunities and how important it will be to ensure that our economy is set up to be focused on foundational technologies so that in the future we can power the economic growth and the jobs of tomorrow. It highlights the fact that the Government are not just making decisions in the short term but in the long term, because, although some of these benefits will come to fruition within the next year or two, for some of them we are talking about how they will impact my son’s future, who is only nine months old, and the type of job that he will do and the life that he will live.
Q4 Lord Hall of Birkenhead: We had one really interesting little session when we went to UCL Business to try to understand AI as it is happening in a university and whether it was getting the right support for spinouts and all of that. We got some evidence that certainly surprised me. I thought it would say, “Well, we could do with more money here or a bit more there”. Actually, it turned out that there has been an alarming drop-off in the number of AI-related PhDs. It was really worried about it. It is something to do with the allocation of funding, I suspect, from UKRI. Is that something that you know anything about or could look into, because obviously our future depends on spinouts and PhDs in AI from places like UCL being absolutely top drawer?
Michelle Donelan MP: Yes. You are talking about centres of doctoral training. While there were some alterations in the approach by UKRI, the actual number of PhD students will be maintained. When it comes to spinouts, we recently did our response to the spinout review that we commissioned, and we announced that in line with the Autumn Statement. We accepted all the recommendations and included a proof of concept fund, because we want to get those great ideas spinning out of our universities.
Lord Hall of Birkenhead: I wonder whether, with the evidence that we got from UCL Business, UKRI is spreading it thin or lots of people are suddenly having ideas like, “I’d better put AI into my PhD application”. There are certainly top-drawer people there, and they were pretty depressed about the fact that they were not getting the money they need for the future.
Michelle Donelan MP: The overall total number of PhD students will be maintained. There were some changes to the way it was allocated, but the overall total quantum will be maintained. We can get UKRI to write with the exact details to the committee if that would be helpful.
Q5 Lord Hall of Birkenhead: They were really good people and very passionate about it and worried about it. They are obviously the future.
Finally, do we need to acquire a sovereign large language model capability? Where, in your mind, do the arguments pro and against lie?
Michelle Donelan MP: One thing that the incubator will be looking at is the best way to make sure that we are deploying AI throughout our public services, whether that requires sovereign capability in large language models, and what will produce the quickest, most efficient and effective results. That question will have to be reviewed constantly over time as the policy emerges. Emran, do you want to add anything?
Emran Mian: The arguments for it would be if we thought that a large language model was particularly suitable for use in certain public services or to allow certain parts of the community to be able to access public services better. At the same time, the large language models that have been developed in the private sector are all advancing very quickly and you can train and tune them very quickly. The reason against it would be that, rather than the state making a massive investment in its own capacity to develop a large language model, it may be better to go along a path where you take an existing model and tune it or train it to specific uses in the public sector.
Lord Hall of Birkenhead: And encouraging it.
Emran Mian: Exactly. We are very keen to make investments that genuinely add value rather than trying to replace investments that the private sector is making.
Lord Hall of Birkenhead: Do you think that the data held in the NHS would be an example of an area where you could see that working?
Emran Mian: That is why we were pleased to see that healthcare was one of the areas that UKRI picked as part of the £80 million spend. As the Secretary of State said in the context of the AI safety summit, we also announced £100 million on AI development specifically pursuing the life sciences missions and the healthcare missions that the Government have set out. Healthcare is definitely a space where we think there is a lot of interesting stuff already going on in the charitable sector and in the private sector, but there is definitely a case for public investment, too.
Lord Hall of Birkenhead: Thank you very much indeed.
The Chair: Just before we move off this topic, may I follow up on a couple of things? On the question of a sovereign large language model, we were clear in our report that this was not something that we were recommending you should do, but something that you should explore further. Of the various ways in which that might be achieved, the route that we thought made the most sense would be for the Government to commission an external developer, a commercial developer, to build it and for it to be deployed on secure government infrastructure and UK-based data processing capabilities itself, both to be able to take advantage of the commercial prowess in this area and their continuing development and to have that strategic resource UK based rather than relying on overseas. That was the critical difference and why we thought it was important to look at that.
To link that point to the question of funding for our research PhD students, I urge you, Secretary of State, to go back to UKRI and explore that and satisfy yourself of the way in which its model of distributing funds is in line with your objectives, because it is not the first time that we have had reason to question UKRI’s approach. The same occurred during our creative industries inquiry. What is happening in practical terms is not necessarily delivering on what you intend.
At UCL, we learned that the risk for a lot of universities and PhD students is that they end up relying on overseas investors, notably China, for funding some of their new innovations. This is an area that we felt was quite difficult and an important place to look at in greater detail and make sure that you really feel that what the Government are deciding will have long-term sustainability.
Michelle Donelan MP: We will get a letter sent to you either from me or from UKRI.
On the point about sovereign capability in large language models, there is a benefit from a national security perspective as well, and we are acutely aware of that. I want to reassure you that the work that the Cabinet Office is doing will examine that exact question. It has recently been onboarding experts just as we have been doing. It is a bit further behind because it has only relatively recently set up the AI incubator in government. It will look at how best to enable the deployment of AI across our public services in a way that is effective, efficient and a good use of taxpayers’ money and that will produce results and bear in mind things like national security.
Q6 Baroness Harding of Winscombe: One of the issues that we looked at in the inquiry was the balance between open and closed large language models. What measures are you taking to ensure that the UK benefits from both open and closed LLMs? Are you taking any specific actions to guard against market dominance?
Michelle Donelan MP: Sometimes the debate can be quite binary on this topic. I agree with the more nuanced approach that you portrayed in your report, which is certainly what we have tried to position in our White Paper consultation response. There are pros and cons to both. We do not need to rehearse them fully, but we all know that open models allow much more access, can advance research and are much more transparent; they can also fall into the hands of bad actors and so on, whereas closed tend to be the opposite.
We need to focus the debate more on potential risks from the models themselves as opposed to just whether they are open or closed. We are currently considering what you could do in this space, and we are in consultation and conversation with developers and civil society. In the White Paper response, we detailed that you need to be much more nuanced and that we need to continue our work in this area to see what the next steps and potential mitigations would be, as opposed to just being pro or anti open or closed models.
Q7 Baroness Harding of Winscombe: One of the risks of closed models is regulatory capture. It is striking that it is the large tech companies that are the keenest on closed models rather than open. How do you respond to the concerns that we set out in our report about potential regulatory capture? What are you doing to guard against that?
Michelle Donelan MP: Taking the question of regulatory capture in a broader frame, if you look at our entire approach to AI, it is not the case that we are only, or even mainly, talking to the big AI companies; we are talking to the entire AI ecosystem, from very small companies to the extremely large, and they were present at the summit. There were some that had only three employees. We are also heavily talking and working with civil society. In fact, we onboarded some members of civil society to the department. One individual came on secondment to help with the response to the AI White Paper.
We have been doing a series of different round tables and discussions, talking to academics and making sure that when we approach our policies they are grounded in evidence. We are making sure that we keep on top of the risks. That is why we set up what we formerly called the state of the science report, which is now renamed. It is basically the report that came out of the summit, where we are getting a full handle and picture of what the scientific research is on AI. It is led by Yoshua Bengio, who is also on our expert advisory panel for the institute. A key part of the institute’s work is research.
In broader terms, we are grounding our approach in evidence. We are making sure that we engage with everybody, not just those with vested interests. That was evident in our summit; its representation was a third from civil society, roughly a third from countries and a third from companies, which was the whole suite rather than just those at the cutting edge.
Q8 Baroness Harding of Winscombe: Do you think that the Digital Markets, Competition and Consumers Bill currently going through this House will help address competition concerns?
Michelle Donelan MP: The Bill is designed to make sure that we promote competition across digital markets. That is the key objective of the Bill, yes.
Baroness Harding of Winscombe: You probably know where I am going. Why, if you think it does, have the Government changed the judicial review appeals standard in the Bill at quite a late stage?
Michelle Donelan MP: It is not abnormal for us to change things. Arguably, we dramatically changed the Online Safety Bill when it had passed many stages. We have listened throughout the Bill. I pride myself on being a Secretary of State who really listens and engages as much as possible with all voices. We wanted to produce a piece of legislation that was balanced and fair, so one of the changes that we made was around proportionality, which I think is a common-sense change. When we subsequently engaged with challenger tech, they were very content with where the piece of legislation is now. In fact, many have outwardly very positively endorsed the Bill that is going through your House.
Baroness Harding of Winscombe: I suspect we will have quite a few conversations about this, because certainly my inbox from challenger tech would not say the same thing. Whether it is the Coalition for App Fairness, which gave us evidence, or a number of challenger technology companies that do not feel able to say anything in public, such is the market dominance of big tech, my inbox would look very different with regard to issues on proportionality, countervailing benefits and the judicial review versus full merits appeal standard.
Michelle Donelan MP: It is a balancing act. I am not for a minute saying that everybody in challenger tech agrees with the exact positioning of where we are, but I would argue that we now have a very balanced piece of potential legislation that will live up to its original policy intentions, which are about promoting competition and making sure that we are standing up for small business as well as big business. Do you want to add anything, Emran?
Emran Mian: The Competition and Markets Authority is using its existing powers to look at some of these very same issues as well. As members of the committee will likely be aware, the Competition and Markets Authority is looking at the Microsoft-OpenAI merger using its existing powers. It is also using its existing powers to look at the cloud market. Our firm view is that the Bill making its way through the House of Lords at the moment will enhance further the powers that the Competition and Markets Authority has to tackle competition issues in digital markets.
The Competition and Markets Authority has already published some work on how it sees the competition issues developing in the foundation model space. We talk about this regularly with the Competition and Markets Authority. Using its existing powers and the powers that build on top of them from the Bill itself, should the Bill make its way into law, we think gives us a really firm foundation for dealing with the competition issues.
Michelle Donelan MP: We did considerable engagement at every single stage of the Bill, and this is the product of us listening. Arguably, we did more engagement with challenger tech. We have been listening throughout.
Q9 Baroness Harding of Winscombe: Engaging and listening are two quite different things when you speak to the challenger tech firms. I would be really interested in how you are thinking about the balance of the Government’s strategic objectives for their relationship with the big technology firms and what you are prioritising. What is more important? Is it good relationships with Silicon Valley so that they bring their international head offices and development teams to the UK, or is it creating an environment that enables UK domestic industries to compete with said big tech companies? How do you think about the balance for the Government’s priorities?
Michelle Donelan MP: I think you are oversimplifying it a bit. We are the department that is responsible for the entire tech ecosystem. It is our job to stand up for competition. At the end of the day, I am a Conservative; I strongly believe in the importance of competition in markets. That is one of the key goals of this piece of legislation. I believe it will achieve that and it is at the heart of the Bill. We want to ensure that we are creating more jobs in the UK. We have one of only three tech sectors in the world that is rated at over $1 trillion. We have a goal within our department to become a science and technology superpower by 2030. We laid out a plan—a blueprint, if you like—as to how we are going to get there, with the science and technology framework. It is incumbent on me and my department to hold the rest of government to account on that as well, with things like skills, et cetera.
I do not think we need to pick one side or the other. We want to make sure that we are growing our own start-ups and retaining them here so that they are not capping out and going to America. At the heart of the speech that I gave just the other week one of my key focuses was scale-up. Equally, we want to attract big business to come here and to locate second offices and international headquarters and have a stake and a footprint in the UK. That is good for our economy as well. I do not think that we need to be binary and choose one over the other.
Q10 Baroness Harding of Winscombe: I am channelling my inner Lord Lipsey, as he is unable to ask his question. I am not hogging out of choice. I am going to change tack on topic and explore a bit more the progress that the AI Safety Institute has been making on assessing risks. Could you start by telling us what progress it has made on assessing risks?
Michelle Donelan MP: The institute has three main roles: first, to conduct research, which we spoke briefly about before; secondly, information sharing, which was one of the things that we agreed at the summit with our international counterparts because that needs to be a joint endeavour; and, thirdly, as we agreed at the summit pre-deployment and post deployment, testing of the models. We are the only country in the world that has managed not only to set up an institute but actually to get it working, off the ground and in operation.
We have seen many countries come forward and say that they too now want to set up institutes. America in the same week of the summit said that. Japan has said similar. The EU is doing an institute of sorts but calling it something slightly different. All of those are taking a lot longer to mature. I call it the Bletchley effect and it highlights the impact that we have had as a country on changing behaviour and actions in this very important agenda.
The institute is already testing the models. That work has begun. We are obviously not providing a running commentary on which particular models it is testing at what point, because some of that information is highly commercially sensitive. We will be publishing an update document to Parliament, on Friday this week, I believe, and that will outline a status update, and talk in broader terms about some of the model analysis the institute has been doing. It will highlight and illuminate a bit more information in the public domain about the demos that it has done and presented at the summit and has presented to certain key stakeholders around the world.
Baroness Harding of Winscombe: Thank you. Sorry, I am going to be boring and come back. I am not sure I understood what that means in terms of the progress on assessing risk that the institute has actually made. Are we able to set out more clearly what we think the big short, medium and long-term risks really are?
Michelle Donelan MP: We did that in the summit, and we continue to do that. We are the only country in the world that has published two documents. One was produced by GO-Science and the other by our security services, and that detailed the current state of play around the risks. We may certainly update that as time goes on, but we did that relatively recently.
Baroness Harding of Winscombe: Part of the reason for my question is that there are two different sets of risks. The committee, as we took the evidence, got quite confused.
Michelle Donelan MP: We put them in different categories. At the summit we categorised them as loss of control, societal harms and misuse. We have changed the names slightly now for the institute because we are trying to stay concurrent with academia and the thinking, and this is an evolving field. Emran might have to correct me if I have got this wrong, but we categorised them still as misuse and societal harm, but we have broken loss of control into autonomous systems, which is a bit broader, and the fourth we have broken up into safeguards. Is it the categorisation that you are after?
Baroness Harding of Winscombe: It is about trying to understand how the AI Safety Institute and the Government are thinking about the different types of risks, and what we should be prioritising in mitigation. One thing from our report was that there is a lot of talk of apocalyptic, catastrophic risk, but we heard time and again in our evidence that it is the more near-term, societal risks that we should be taking more seriously. I am trying to understand where the AI institute has got to in assessing which risks it should be acting on now, versus doing research that might touch your and my children’s lives in 50 years’ time.
Michelle Donelan MP: I think that is the wrong positioning. Actually, we do not know when some of these risks may materialise. Some of the risks could be a year off, two years off, or two months off, because the next set of models is coming out. Some of the risks are already existent. When we talk about societal harms, we think about bias, discrimination and the potential impact on elections. They are very live and front and centre in our minds. It does not mean that, equally, we should not be prioritising things like misuse, autonomous systems or safeguarding. This technology is developing quicker than any technology that we have ever seen. When you speak to the developers, there is a great deal of debate over the pace at which the technology will develop.
One of the key things the Prime Minister set out in his speech ahead of the summit was that we have to be responsible as a Government on this. We cannot just sit back and say, “Well, let’s only do the risks that we can see now”, because in literally two months’ time, four months’ time or a year’s time, if the other risks materialise, we will not be prepared. What we are doing is research across the board of those categories of risk. We are utilising the institute to do that, but we are not just relying on the institute. We have also set up our scientific report, which is the collation of research across the globe by Yoshua Bengio, and is one of the key outcomes of the summit.
We are working with our international counterparts. One of the announcements that we made in the White Paper was £9 million where we are working with the US, trying to stay on top of the risks and making sure that we maximise our expertise, working with our allies. I do not think it is a case of us saying, “Okay, as a Government let’s pick this category of risk and prioritise it over the others”. I think that would be irresponsible. We need to be prioritising all the categories of risk so that when the institute examines the models, it is not just looking at one area of risk but is looking at societal harms as well as some of the other risks.
Baroness Harding of Winscombe: For the near-term risks that you mentioned—misinformation, disinformation and the exacerbation of existing harm, whether that is terrorism, cyber or disinformation—have the Government scaled up mitigation?
Michelle Donelan MP: Yes. Our approach is to make sure that, as the institute examines the models that are coming forward, any of those risks are identified so that the developers can fix them. That is a mitigation. When we talk about some of the other areas—for instance, disinformation—there is no silver bullet when it comes to some of the potential risks. If we think about the use of AI potentially in a general election, we are taking that very seriously, as you would expect us to do. What we are doing on that is to have an approach led by the Home Office called the Defending Democracy Taskforce. We are working up potential solutions and mitigations. Some of them lie within the AI itself, such as AI detection technology. There are then various debates over whether that can be jail-broken or not, or watermarked.
Meta just made an announcement today. We are waiting on a briefing on that this week to learn how effective they believe it will be. We are staying very close to the technology. Some of these things do not have one simple solution. Some of them require a collection of different solutions. We need to keep progressing our knowledge in these areas to advance potential mitigations.
Baroness Harding of Winscombe: Have the government agencies charged with countering these harms seen their budgets increase as a result of the risk that AI brings being identified?
Michelle Donelan MP: Some of that is borne by the regulators. The whole point of our White Paper is to say that it is not the case that all this stuff is a wild west free-for-all. We have a plethora of regulation and legislation in this country that interplays in many different respects, and a group of regulators that are world renowned for their approach.
One thing at the heart of the response is very much saying that we need to make sure that they have the support and the skills to be able to approach this new and emerging area. We are dedicating the £10 million that we allocated in the announcement. We are also utilising our central risk function to support and help. We will be horizon-scanning for potential risks, and we are constantly evaluating to make sure that, if there are gaps in our current regulatory framework, we can look at how we would fill those gaps. We have set up a steering group to that effect.
It is not just about how we could create new; it is about how we can utilise the existing framework to make sure that regulators are working to the correct principles that we outlined in the White Paper, and that they have the support and skills to be able to do that as well.
Baroness Harding of Winscombe: Forgive me, but that sounded like no, they do not have more money.
Michelle Donelan MP: They have £10 million for skills and support. Many of them have had recent budget settlements as well. For instance, ICO is within my remit, where there is an ongoing process. They are individual decisions for each department that is responsible for their regulator. In the process that I outlined, we are going to identify whether there are any holes, gaps or barriers. Should that require additional money, we would be able to support in that regard. Should it require legislative changes, we would have to go down that road. We are very open and honest in the document about how we are making sure that we stay on top of exactly what our regulators need to be able to fulfil this.
Baroness Harding of Winscombe: Thank you. I am mindful that I am hogging.
The Chair: We sent a letter after we had finished taking evidence asking for further information on the resourcing for the central function of how you were dealing with the near-term risks. Unfortunately, we did not get any further information from the Government. One of the reasons why we were keen to understand that was because, contrary to what you have just said about the regulators, we would expect things like disinformation, child sex abuse, counterterrorism and security, to be managed centrally within government rather than just farmed out to individual regulators.
Michelle Donelan MP: We are certainly not saying that we are farming out policy decisions or approaches to regulators. The point I was making was that we have a suite of regulators already whose job is to regulate certain sectors, and therefore to regulate the interplay of AI within those sectors. When they were originally established, that was not their daily business. We now need to make sure that they have the skills, the know-how and the support, and that there are no gaps, to enable them to do that. When it comes to policy decisions or some of the topics that you mentioned, of course we would do that centrally. The central risk functions remit is a unit in my department.
We have not gone down the road of setting up an AI regulator. It would not make much sense, especially at this stage, because our approach is context-driven and you would need experts in each area. To make sure that there are no gaps in the co-ordination of the regulators or between the differing approaches, we have set up a unit in my department to assist in that. That is the way forward. Emran, do you want to add anything?
Emran Mian: As well as expanding capacity, we have to study safety at the foundation model level, which is what the AI Safety Institute is focused on. We have been growing the team that looks at wider issues of AI policy, both the opportunities as well as the risks. It is in that team, in the AI policy directorate, which has also been growing significantly, that we have the central AI risk function that the Secretary of State has been describing. That is where we focus effort on looking at the deployment risks.
We resist the terminology of near-term or long-term risks, because there is a chance that a lot of the risks that are associated with the frontier models could manifest in the near term rather than in the long term. Monitoring of the deployment level risks, whether disinformation or the use of AI to drive financial crime—those kinds of issues—is a function that we have grown within the department as part of the AI policy directorate exactly because we need one place in government that takes a coherent view of what that looks like across all the different use cases or the different sectors and different regulators.
Where we identify that there are issues and we need to go further, we have been very open in the document published today that we are ready to fund enhancements and capability. For the moment, we have put aside £10 million to do that. We are very happy to look at where there may need to be further context-dependent legislation or regulation.
Michelle Donelan MP: It is born of need as well. That is why we are supporting the regulators, rather than just giving them a load of cash. We need to identify what the money needs to be used on, and where, to be effective for taxpayers and for this agenda.
The Chair: Notwithstanding that you do not use the term, if you can provide us with more information about how you have scaled up that central function to deal with what we are describing as near-term risks, as in risks that exist but are becoming much more magnified by the ease of access people have to make them real, that would be helpful to know. You can follow up on that in writing. You do not need to do it now.
Michelle Donelan MP: I think we can illuminate it more. On our actual manpower on the AI agenda, as Emran says, we have been dramatically onboarding people over last year. The level of expertise we now have in the department in this area means that we can get underneath the bonnet, stay on top and stay current with the risks and opportunities.
Q11 Lord Young of Norwood Green: Good luck with this one. You are certainly trying to cover the waterfront, which is obvious, bearing in mind that it was not that long ago that the doomsters said, “Well, we’ve got to pause everything for six months”. I thought that was impossible and was never going to happen. We are not doing that.
I am interested in two things. First, obviously there is the role of regulators and government, and you setting up your department to cover where the risks are. There is also the importance of individual companies recognising that they are at risk. We have seen what happens when they have been penetrated. It is quite interesting. You ignore cybersecurity at your absolute peril. You have to get a balance between what you do and ensuring that companies are focused on that.
The last point is that while we have been focusing on risk, and rightly so, what we are seeing is that this presents a dramatic opportunity for—
The Chair: We covered that earlier on, Lord Young.
Lord Young of Norwood Green: Maybe we have, but I still want to put the question. It is getting the balance right on risk and ensuring that you are also encouraging people to use the technology and make the most of it. Do you think that you have the balance right? Are we generating enough skills on the cybersecurity front, both for you to use and for companies to use?
The Chair: In the interests of time—forgive me, but we dealt with the balance between opportunities and risk earlier before you joined us, and we are coming on to skills later—unless there is something that is not covered by either of those things that you want to add in response.
Michelle Donelan MP: We can hoover it up later, if you want.
The Chair: We are coming on to skills a bit later, so we can pick up on that there. Baroness Wheatcroft will talk about copyright.
Q12 Baroness Wheatcroft: Thank you. I declare an interest as I have an involvement with the Financial Times. The Financial Times is one of the organisations that gave evidence to this committee fearful that its intellectual property was being stolen by big tech. At that stage, government policy was to try to persuade big tech to behave properly. From the White Paper today, it seems that you have changed your mind on that. Could you elaborate a bit?
Michelle Donelan MP: We obviously want to create an economy that is fostering innovation in AI. That is certainly the path that we have been pursuing, and I would argue successfully. Secondly, we of course want to support the creative sector, which is one of the world’s leading creative sectors. It is extremely important not just for our economy but for communities up and down the country. Previously to this, of course, I was Secretary of State for DCMS and saw at first hand its added value and importance. Equally, as a constituency MP, I see it locally. We are trying to do both.
We are not the only country that wants to try to achieve both of these things. AI, as we have said many times today, is an emerging technology. The issue has developed relatively quickly, so we tried to bring both sectors together, working with the IPO, to see if we could achieve some commonality and produce a code of conduct. As we are really honest in the document published today, that proved very difficult. We did not manage to achieve a code of conduct. There was commonality in some areas, particularly on transparency. I have been talking to my international counterparts in many countries around the globe. They, too, are struggling with the path forward on the topic, and we have been trying to get some international cohesion and synergy.
We are saying that we will continue working with both sectors—both the Secretary of State for the Department of Culture, Media and Sport and me. I spoke to her about this very topic this morning. We will continue that work. We think that we have a path forward that will focus particularly on transparency. We do not want to rush to announce something that damages either of those sectors. Even if it means taking a little bit longer to get it right, we think that is the best approach.
Baroness Wheatcroft: But in focusing on transparency, potentially, are you ruling out actually legislating to protect the copyright of content providers?
Michelle Donelan MP: We are certainly not ruling out legislation. We may well have to legislate eventually to achieve the desired outcome. We are still continuing to talk to all stakeholders. As I said, we have a pathway forward, but we need to do more engagement now with our stakeholders to make sure that that would work for both incredibly important sectors.
Baroness Wheatcroft: One of the things that Viscount Camrose mentioned when he gave evidence to the committee was that the big tech providers simply have the option of doing their mining from other locations outside this country. Can you say how you feel the Government might be able to contend with that in protecting the copyright interests of organisations within this country?
Michelle Donelan MP: I think that highlights why it is important that we do not just look at these topics domestically and that we work with our international counterparts, in particular America. This is certainly a conversation point that I have for my next meeting with Secretary Raimondo. It is a topic that I have discussed with many of my counterparts across the globe. You are right: we live in an interconnected world. The creative sector especially is, in essence, a global sector. It is important that we do not just look at it exclusively from a domestic perspective.
Q13 Baroness Wheatcroft: You said that during the IPO discussions there was some commonality of interest. Could you say a little bit about the areas where there was not commonality of interest and where, in fact, the big tech companies refused to play ball?
Michelle Donelan MP: I am not sure it is helpful to keep calling them big tech. We are talking about the entire AI sector, some of which are start-ups. When we talk about big tech, it sounds like we are talking about very large, predominantly American, companies, when we are actually talking about the entire AI sector where we want to foster innovation and get more spinouts, as we spoke about before, from our universities. We want that innovation, because it will mean that we can have the opportunities, as well as the jobs and the benefit to our economy.
The disagreement was obviously that the two sectors are coming from different perspectives on what would be beneficial for their sector. We think we have found some commonality and a way forward that is different from the code approach. Would everybody be happy and over the moon? No, but we think that we can get to a position that would deliver. We need to continue that engagement now.
Baroness Wheatcroft: Would you accept that in the meantime copyright is being infringed by organisations, both big and small, that are mining for their databases? How long do you think you should be prepared to tolerate that?
Michelle Donelan MP: Do I think that this is having a real impact and that our creative sector is concerned? Yes, I do, and I know that they have made those representations. Equally, do I think that it is important to protect the AI sector and stand up for innovation? Yes, I do. We cannot rush and get this wrong. We have tried. It is not the case that we have done nothing. In fact, we are ahead of many of my international counterparts in our thinking, our policy engagement and development in this area because we recognise how important it is. It is not a case of us just sitting back and waiting. It is about us finding the best way forward. A lot of the stakeholders I have talked to certainly would not want us to rush this and get it wrong, potentially damaging their sector.
Baroness Wheatcroft: I suspect there are a lot of content providers who would be very grateful if you rushed in, because they are losing money all the time.
Michelle Donelan MP: Not if we got it wrong.
The Chair: But you think the copyright principles still stand.
Michelle Donelan MP: I think we need to find a solution that works for both sectors. I do not want to be drawn on exactly the solution we are working on, because we have not made that announcement, and we think it is really important that we take people with us on this and engage with both sectors. It is also an area that is co-owned between me and DCMS, so I do not want to be drawn any further on that.
The Chair: We move on to digital exclusion, but as part of that we will still draw on AI-type questions.
Q14 The Lord Bishop of Leeds: I feel that this is a bit of delayed gratification, because we are going to get to skills once we have addressed this. As our two inquiries have gone on, it has puzzled us as to why the Government have maintained that digital exclusion is a priority while at the same time refusing to update a strategy that is a decade old.
Michelle Donelan MP: I totally understand the point that the committee is making. The principles that that strategy outlined still exist. If we were to renew and redo that strategy, it would chew up a lot of government time and resource, whereas my approach is more focused on delivery and making sure that we can actually do the job in hand. We have tried to address it in a different way.
For instance, as we said in responding to your report, we have set up the interministerial group, which is reviewing across the board, across all the different departments. It is almost a stocktake of where the potential gaps are and the things that need to be addressed more in terms of digital inclusion. I have been making sure that it will look not just at the here and now but at the future, and project where we foresee the emergence of technology in five or 10 years and the problems that will present, and where we need to be plugging some of the holes now and what we need to be doing now to prepare us for the future. That is one of the key pillars of our work.
The Lord Bishop of Leeds: There are lots of areas of government activity where the principles still hold but strategies are updated. The strategy provides the plumbing for how we are going to achieve what we want to achieve. You referred to the interministerial group. Can you flesh that out a bit more? What has it achieved in the time that it has been in operation? The principle of it obviously makes sense, but what difference has it made?
Michelle Donelan MP: This is an area that is, in effect, owned across government, as you know. DfE has a stake in it, as does DWP. There are various initiatives across the board that are helping to target digital exclusion and promote digital inclusion.
The key purpose and remit of the group is twofold. First, it is to raise the importance, not just in our department but in every department, of the priority that is placed on this area, to make sure that it is not lost in everything else that is happening. Secondly, it is to make sure that each department comes forward with a thorough stocktake on what they are doing on this agenda, to see if there are any gaps or areas that need to be accelerated or advanced.
There is also future-proofing by looking ahead, as I said a moment ago, to what we need to be doing in a year’s time, two years’ time or now in laying the foundations for where this might be going. There has been one meeting already. That work has been commissioned by each department. It is my understanding that that is currently the process that is being undertaken ahead of the next meeting.
The Lord Bishop of Leeds: Thank you. One of the interesting things that came out of the inquiry was the recognition, which you have alluded to, that, because it is cross-departmental, many of the Government’s objectives such as levelling up and improving public services would benefit from a joined-up, strategic approach. That is why I am a little concerned about what the timelines are for making a difference, for identifying the gaps and addressing them.
Michelle Donelan MP: I do not have the date to hand, but we can certainly check when the next meeting is and write back to you on that. We can also keep you up to date with the work of the group.
You said a moment ago that a strategy would provide the plumbing for this piece of work. I believe that actions speak louder than words. If we can achieve this in the here and now, and more quickly than spending the next four or five months writing a strategy, that is the approach that we should take. Let us not forget that there is a great deal of work going on in the area of digital inclusion throughout every department each week. We can bring that together and articulate it, to make sure that we focus on it, without necessarily packaging it up in a strategy document.
The Lord Bishop of Leeds: Our inquiry revealed that there are a lot of areas where we hear that things are being addressed, but we do not necessarily see the outcome. I would be interested to know how you respond to Liz Williams, CEO of the business coalition FutureDotNow, for example. She put it to us that the Government were not taking digital exclusion seriously. Her argument was that, “It’s everybody’s and nobody’s responsibility at the moment. We don’t have a clear national ambition”. In other words, because it is a dispersed area of interest and activity, there is no coherent approach to it.
Michelle Donelan MP: I disagree. There are many areas where each government department has a role to play or a stake. That does not mean that it is not a priority. In fact, you could argue the opposite. You could say that because it is not just being worked on by one department, but is getting 10 working on it, it is even more important.
The Chair: We are still in this area, but we move to Baroness Healy.
Q15 Baroness Healy of Primrose Hill: Secretary of State, generally, what progress have you made on supporting digital skills for all sectors of society? There are millions still excluded, both the young in schools, as we found during the pandemic, and the older population who cannot use these services. Obviously, there is the interministerial group now and the Cabinet Office work, but what do you think are the key things that you can do to address this very urgent problem?
Michelle Donelan MP: I basically see it as three areas. First, you have what I call the skills for life. You have to be able to have the digital skills to access products and commodities in the modern world, whether that is buying your parking in a smart way, using online banking or using your computer for connectivity. The second area is skills for employment. The third area is skills if you wanted employment in actual digital roles, where the level of your aptitude and qualifications would be higher.
The first of those is certainly something that many government departments are feeding into. There is the digital skills entitlement, which is a policy run from the Department for Education in utilising community learning. There are a number of initiatives by the Department for Work and Pensions, especially for those who are claiming universal credit to help them get into employment. In my department we have schemes such as Cyber Explorers, so that we can ensure that even young children learn about the risks of cybersecurity, so that they grow up acquiring skills that you need in the modern age that would not necessarily have been as fundamental in previous decades.
On skills for employment, the Department for Education has quadrupled the number of bootcamps. I think practically all of those have a digital element. We have our Digital Skills Council, where we have been listening to industry and better understanding the needs that are required. We have tried to shape our education system to accommodate modern employment needs, through things like institutes of technology, which were established in the Department for Education, in fact, when I was a Minister there. T-levels have been rolled out, and the LLE, which we will launch in 2025, will enable people to study up to four years of degree-level study throughout their life in modular, bite-sized chunks. That will enable people to reskill and upskill much more easily, including in areas that involve digital aspects.
The next tier is individuals who want to seek employment within the digital sector and who will need a higher level of skills. We have already spoken about the centres of doctoral training and AI PhDs, for instance. We have also invested in master’s conversion courses for those who have not studied STEM qualifications before. We have given 2,600 individual scholarships, particularly for those in underrepresented backgrounds from a financial perspective.
Just last week, when I was in Canada, we launched an AI exchange programme. I am choosing some highlighted examples to show you that we are trying to look at all the different layers. At the moment, we talk about digital exclusion, but most people would talk about it from more of a basic skills perspective. When you think about what digital exclusion and the definition may entail in 10 years’ time, it will be remarkably different. That is why I think it is important to focus not just on basic digital skills but on all the layers.
Q16 Lord Young of Norwood Green: I have a number of issues I want to raise. There has been a lot of criticism from employers about the apprentice levy needing to be revised and that it is not flexible enough to meet some of their needs. That is one issue.
At the moment, you have the challenge that the number of young people in the 16-24 age range has actually dropped. Encouraging SMEs to take on apprenticeships is another challenge. You need them to do that. It is not good enough for the large companies to do it. You have to get them to do it. You have had problems in the past with transferring levy money and stuff like that.
I have given you a range of issues that you and the Department for Education, et cetera, have to address if you want to deliver the kinds of skills that you are trying to encourage in AI, whether it is cybersecurity or whatever. Do you think that you are on top of it? I apologise for raising a number of issues.
Michelle Donelan MP: That is okay. It is very apt in National Apprenticeship Week as we celebrate the vast opportunities that are available in this country with apprenticeships. One thing that I think we have done very successfully in this country is remove the stigma associated with apprenticeships and highlighted how they can offer a pathway on a par with, if not better than, some of the traditional routes.
I held a careers fair in my constituency the other week. What was striking was that there was a mixture of people of all ages. The feedback from every employer there was that the vast majority of people were asking about apprenticeships. If we think back 10 years ago, I do not think that would have been the feedback.
Lord Young of Norwood Green: I absolutely agree with you. Yes.
Michelle Donelan MP: We have a great story to tell. We are one of only four countries in the world that offers apprenticeships to all ages. We have a very successful system of apprenticeships in the vast majority of career occupations. The quality of those apprenticeships has been dramatically increased and improved over the last few years.
We are not about to make the levy system more flexible. In fact, the plans by the Labour Party to do so would result in halving the number of apprenticeships available. One of the requests from industry is to make it more flexible so that they can offer alternative training courses. The direct result of that would be to reduce the number of actual apprenticeships available. At the moment, there is a transfer of the levy from bigger businesses to the smaller businesses that you referenced. Obviously, if a smaller amount of that transfer was happening, fewer small businesses would be offering apprenticeships. We need to stick with our current plan on apprenticeships. It is not only working but delivering, not just for our young but for people of all ages and stages across the board in different career sectors.
Lord Young of Norwood Green: Do you really think that you are okay on every front? Some of the statistics do not support what you are saying when it comes to SMEs and the number of young people in the 16-24 age range. Put to one side whether you should or should not reform the levy. I do not necessarily agree with it, but I accept the point you are making. That is not to ignore what you have done overall. I am not criticising that; I am a supporter of it. I am a national apprenticeship ambassador, so I am trying to encourage them, but the feedback I get from SMEs, for example, is that the administration is too complicated and deters them from taking part. I think you should be addressing that. You see it reflected in those statistics. I am surprised that you think everything is okay.
Michelle Donelan MP: What I was saying was that we have a world-renowned apprenticeship system that is revered by many countries around the world. We have managed to be the architects of that system over the last 10 years, supporting businesses of all sizes and people at all stages and ages. Many of my own constituents have benefited from that structure.
The Department for Education will continue to listen to and engage with industry, colleges, universities and schools to make sure that we are constantly improving not only our approach to apprenticeships but our communications about apprenticeships. One of the big changes that we have made within the last year is adding apprenticeships to the UCAS system so that there is complete parity. That not only removes a last remaining element of stigma but makes the system easier for people to access.
Lord Young of Norwood Green: Thank you.
The Chair: I have a comment, and then one final question. We as a committee are hugely ambitious for the UK on AI and the wider technology sector. We are ambitious and very optimistic. We believe that we have the talent—all those sorts of things—but we are concerned not to leave anybody behind and that everybody can take advantage of this technology and has the opportunity to be a part of it. That is why, for us, the question of digital exclusion is a central part of the whole agenda.
I think you are right to say that how we define digital exclusion will develop and change, which was one of the key messages of our inquiry. This is a moving target, and the concept we had as a political class 10 years ago when we talked about digital exclusion is very out of date. It is quite different now. We discovered that even if you think that you are digitally included at one point, you might suddenly become digitally excluded because you have not kept pace with the technology.
We would urge you, Secretary of State, in the context of your role and your department, to think about how, if you do not want to renew or refresh the strategy, you can make sure that digital exclusion is brought much more front and centre and is a bigger part of your bigger agenda in terms of innovation and technology.
Michelle Donelan MP: I can completely agree with you. There is absolutely no point gripping all these risks and promoting innovation if we are not going to ensure that we get adoption across our businesses, our public services and our general public. We want people to seize the benefits of this incredible technology and technological change. We want to make sure that we are at the forefront of that as a nation. That is why digital inclusion is particularly important. We need to do more research into the potential barriers and the reasons behind digital exclusion. That is something that we will pursue.
The other thing that I want us to pursue, which we will, is more research on the interconnection between digital exclusion and AI. We have to remember that, as AI is emerging so quickly, this is another area that we need to stay on top of. Some of it is anticipating how AI will develop and the impact it will have on our society, our way of life and our employment patterns, and then to predict and anticipate the impact it will have around digital inclusion. That is why I think we should be doing more research on that topic, which we will be commissioning.
Q17 The Chair: That is good to hear. My final, very small, question is this. You referenced the Cabinet Office and its AI incubator in today’s session. Within government, who is in charge of AI? Is it DSIT or the Cabinet Office?
Michelle Donelan MP: I am not sure that it is quite as simple as that. We are the department of technology, and we are the lead department when it comes to AI. The AI Safety Institute sits within my department, and it is accountable to me and the Prime Minister. We are the department that established and ran the safety summit just a few months ago. We are working with the Republic of Korea on the next one, and with France on the one after that. We have all the work in train around AI safety and ensuring that we boost adoption across our businesses.
Public sector procurement has always been done out of the Cabinet Office. It makes sense, therefore, that it is the lead department looking at how that procurement needs to change, adapt and evolve with the interplay of AI. That said, we work very closely together. I had a bilateral meeting with the Deputy Prime Minister on that very topic just a few weeks ago. We are acting in an interconnected fashion. You could argue that AI will impact every department. Some of the most exciting use cases that we have talked about today when it comes to the general public’s reaction and impact are about health. We work very closely with the Department of Health and Social Care and with the Department for Education, but the one department that is the conduit and the driving force is our department.
The Chair: Thank you again, Secretary of State, for your time and for appearing before us today. Thank you, too, Mr Mian. It is good to have you with us.
