Uncorrected oral evidence: Multilateralism
Oral evidence, 28 October 2026.
The Chair
Welcome to the fifth public evidence session of the International Relations and Defence Committee of the House of Lords looking at the multilateral governance of artificial intelligence. It is being livestreamed on parliamentlive.tv, and there are people watching it. The transcript of today’s session will be sent to you, so any minor modifications or corrections that you might want to make can be delivered to us. I ask any members asking questions to declare any interest they might have that is relevant to the subject. I declare that I am a senior counsellor with the Cohen Group in Washington. When you answer the first question, please introduce yourself to the committee and the outside world. All the initiatives that we have seen so far in AI governance have involved Western powers and China. There is an argument that only those powers with the technological advancement and knowledge should set the norms and standards for AI, and that they be adopted on a global basis. Can you comment on that and what the potential strengths and weaknesses are in the concept of the existing powers setting the rules for the future?
Dr Rachel Adams
Good morning and thank you so much for having me. I am a research professor at the University of Cambridge, where I lead the Leverhulme Centre for the Future of Intelligence, which I believe is the world’s oldest research centre on AI and society. It was established about 11 years ago. I recently took up that post. Prior to that, I was living in South Africa for 18 years, working very closely with the South African Government and Governments across the African continent. Three years ago, I founded a think tank called the Global Center on AI Governance, which is headquartered in South Africa but has a mandate to work across global majority countries, supporting South-South co-operation on AI and particularly helping ‘Global South’ Governments to establish governance norms that are pertinent, applicable, relevant and suited to their own contexts and domestic priorities, while thinking about all these issues within an international context. Chair, I will start with your question about who is setting the norms around these technologies. Of course, that is a very emerging area. We do not have many international frameworks or binding laws that constrain how AI is developed and used around the world. There are a number of problems with that. One is that countries that are not part of those arrangements and norm setting will receive norms, laws and standards that are either not useful to them or that undermine what they might be trying to do. We have seen examples of this before, with the impact of GDPR internationally. In setting a very high standard for the sharing of personal data globally, GDPR has, for example, restricted international research collaborations that have involved the sharing of genomic data. In the longer-term view, if there are only a certain set of countries that are setting the standards and laws around such a strategically important technology, we are going to create international stability by leaving behind parts of the world that are rapidly taking up newer positions of power. We are in a multipolar world, and it is not one where a single world order is dictating how the rest of the countries might engage internationally. It is very short-sighted to think only that countries such as the US are setting standards that are applicable for everybody. The question is also partly moot because the US is pushing for deregulation, but what we really need is stricter governance, enforcement mechanisms and institutional capacity to govern everywhere in the world, because this technology can take us in many different ways; some that might be hugely beneficial and some that might be very risky and could undermine values that we have spent the large part of the 20th century collectively building.
Baroness Fraser of Craigmaddie
I was interested to read in our briefing notes that when South Africa held the presidency of the G20, it commissioned research that said that less than 10% of all global governance frameworks on AI have emerged from developing countries. You have just articulated why that is not a good thing. Do they themselves feel left out of this debate? What issues and agendas would they want to see as most important? Since this inquiry is looking at the role of multilateral agencies in this space, what mechanisms, structures or organisations exist for such issues and agendas from developing countries to be fed in?
Dr Rachel Adams
I will answer that in three parts. First, what is the status of AI governance in the ‘Global South’? Secondly, what are the issues that ‘Global South’ Governments care about? Thirdly, what international mechanisms exist? To answer that first question, I think that we can update those statistics. I was very involved with South Africa’s G20 presidency and there is a lot that we can discuss about how the UK’s G20 presidency can pick up some of the consensus that South Africa built. The Global Center on AI Governance, which I lead, conducts a study called the Global Index on Responsible AI that, interestingly, is funded by the FCDO. That index measures what 138 countries around the world are doing to develop governance frameworks on AI and implement them. What we have seen over the last two years is that the ‘Global South’ has led the charge in developing AI governance frameworks. If you look at the total number of governance frameworks on responsible AI that have been published and proliferated over the last two years, over half of them come from ‘Global South’ countries. However, 78% of those frameworks are non-binding. They either do not contain any provisions to support implementation or they are not enforceable and do not provide protections for individuals or communities at risk. What they do show is positive executive intent. They show the hyperscalers that they are thinking strategically about the value of AI for their countries and that they are open to the kinds of partnerships that are flowing in, but they are not ensuring that there are safeguards in place to protect human rights, ensure democratic accountability and so forth. We are also seeing an implementation gap in the frameworks that exist in the so-called global North vis-à-vis those that exist in the ‘Global South’. The statistics are something like 45% of all the frameworks that exist in the ‘Global South’ are being implemented, whereas it is 65% in the global North. You have to caveat that with the fact that many of these frameworks are new, but we are also looking for whether they contain provisions that would enable implementation, such as oversight authority authorities or budget provisions that would enable budget to be put forward to support their implementation. Where I see huge difference is in the articulation of what the risks are. This is probably the starting point for why global governance around AI is so fragmented—we do not have a shared language for what the risks are. In the global North, we are talking a lot about the existential risks of the dual-use problems associated with these advanced technologies and the frontier risks and how we get access to frontier capabilities in order to test and evaluate them. But in other parts of the world, the risks are quite different. But, in other parts of the world, the risks are quite different—the enforcement capabilities and a lack of institutional capacity to manage the risks that might be coming with them. In South Africa, for example, data centres are being established, and a number of people are making complaints to the South African Human Rights Commission about human rights concerns they have with these data centres being established. The commission is now overwhelmed with more cases than it is able to respond to. There are really big risks around labour. It is not just about labour displacement, the impacts on job markets in terms of jobs disappearing and the shared risk—this is a really important issue in the UK, too; in parts of the ‘Global South’, it is accompanied by the new forms of labour that are occurring through the AI supply chains and value chains. For example, there is data-labelling work, where people tick, “I am not a robot”. For many months on end, their work is just consumed with this digital micro-labour to manage and help train AI systems. This kind of work is very badly paid. There are no labour protections associated with it. There has been significant whistleblowing around this. Of course, if there are no labour protections, accompanied by weaker institutions that are not able to enforce protections that might exist or apply laws in new areas, it creates a real vacuum. Some of the other concerns are around the environmental impacts, which are unevenly distributed and very hard to attribute. There is an increase in environmental disasters, largely in the global North. We do not know quite how far AI is contributing to it, but that is another area that people are certainly concerned about. Another area is public understanding of and public literacy on AI. In South Africa, we did a study to assess what people knew and perceived about AI. We interviewed over 3,000 South Africans of all walks of life in their local language, and almost two-thirds of South Africans did not know what AI was. Those who did know got their information through algorithmically mediated platforms. Again, there is a democratic gap in Governments moving forward with trying to adopt AI, particularly in the public sector, and people being able to understand what is happening and being able to challenge some of those decisions. A lot of the very real risks that are being felt across the ‘Global South’ are much more immediate than some of these more hypothetical concerns about existential risks. There is a shared set of risks around things such as AI sovereignty and the dependence on foreign tech providers, particularly for critical services—whether that is defence, national health systems or social grants. That risk is shared globally. I would think about how the UK can play an important role in supporting ‘Global South’ Governments to negotiate stronger deals and terms with the hyperscalers that come in, particularly in the absence of binding domestic frameworks. I am happy to go into more detail on anything else I have spoken about, but the third question was on what international mechanisms exist. The most important one, from a few years ago, was what UNESCO had developed. It developed the ethical framework for AI and spent time evolving ‘Global South’ countries in developing the principles and values that are articulated there. When it was promulgated, there was a shared sense that, “Yes, this also reflects our interests”. Then, UNESCO developed its readiness assessment tool, which provided countries with a mechanism for seeing where the gaps were in relation to this international standard and where their own regulatory regimes were at. Our centre led that work in nine southern African countries. It was very difficult to do it, because the questions asked referred to global indices that simply did not include those countries, such as broadband, gender-related issues or representation in Parliament. There was a range of different things, but we do not have adequate, globally representative evidence from all these different countries to be able to support some of those internationally led processes that are designed as one-size-fits-all. More recently, we had the UN Global Dialogue. That has been important, I so far as it has been very deliberately inclusive of countries everywhere and multi-stakeholder, but of course there are thousands of voices all with two minutes to make a short statement, and there is very little room to listen to one another and figure out the areas of shared concern and consensus. The UN has a very tall task ahead of it. There will be the May dialogues, and I imagine that this big mandate will devolve into a set of thematic issues and the countries that care most about some of these thematic issues will position themselves more strongly on AI safety and security, access to compute and compute infrastructure, or talent and capabilities. I worry about how that may move forward. China led the World Artificial Intelligence Cooperation Organization—WAICO—which was launched in July this year. We should talk about that. It has not gone very far, certainly not far enough to say how influential and impactful it will be, but it was important in so far as it articulated a starting point that was relevant and pertinent to ‘Global South’ countries on how they could have a political voice globally. The developmental focus in terms of AI supporting socioeconomic development was front and centre. It also articulated issues around training and capacity strengthening. It is important to see where that might go. Ten African countries have signed on to that. There have been a couple of collaborative efforts since to send African diplomats to China for training. It is important to look at what they centred their emphasis on and why that reflected the interests and priorities that were important in the ‘Global South’. The G20 was particularly valuable, and it is unfortunate that South Africa was not able to pass on to the US in the way that that should have unfolded. That was an important moment for South Africa to represent the broader ‘Global South’ and the broader African continent and put forward concerns around the labour impact, unequal access to compute infrastructure and the need for governance standards that reflect diverse priorities. That was important for putting some of those things on the table. Q31 Baroness Blackstone: Much of the commentary about middle-income and low-income countries and AI has focused on these countries being a battleground for Chinese and US interests. Can you comment on that and in doing so, say what the advantages and disadvantages of that are? Dr Rachel Adams: My perspective on this is that this is a Western framing that is not how African Governments see what is happening strategically in their jurisdictions. I do not think they see themselves as a battleground for US and Chinese influence. It is important to say that the US influence is now almost solely through the companies whereas, previously, USAID had enormous influence and relevance in shaping developmental agendas across the ‘Global South’. That now does not exist. We see the FCDO and others, such as Sida in Sweden and the Swiss Agency for Development and Cooperation, playing important roles there. From an African perspective, it does not feel like a US-China dichotomy. The US hyperscalers are coming in with very big offers to develop AI infrastructure and cloud computing and establish data centres. These have been hit by national constraints. In Kenya, a massive data centre build was not able to move forward because there simply was not the energy infrastructure to enable it to be established. Certainly, those of us who work on the human rights and democratic accountability side are concerned about these hyperscalers coming in without strong negotiations and contractual terms that provide local Governments with the opportunity to leave a partnership or change a provider, should they want to. Countries should have a diverse range of foreign tech providers, where they are relying on foreign tech providers. That is much harder in resource-constrained environments, but it is important for issues of sovereignty, which is not solely about ownership but control, management and governance. China has an important influence in the offer of these open-source or open-weight technologies. There are various degrees to which the technologies of Qwen and DeepSeek are completely open, but they are more accessible to African innovators, who are able to adapt them for local use. There is a lot of work in adapting these technologies to work in local languages and work on local issues. I am not an expert on the open-closed conversation to be able to comment on that. The point I am trying to make is that it is a Western framing on the US and China. Strategically, if you are engaging with African countries and with ‘Global South’ countries, that is not the framing to use; instead, you should be thinking about what the practical and helpful partnerships are that can support these countries to have their own agency and to participate, and be allies, in international co-operation on AI. Q32 Lord Darroch of Kew: Dr Adams, there is a compelling statistic in our papers that, in 2025, Microsoft found that 25% of the working population of the global North use AI in the workplace, but only 14% do so in the ‘Global South’. Can you give us your judgment on how successful multilateral, international government initiatives on AI have been in spreading technological capability to low-income and middle-income countries? Dr Rachel Adams: I think that was Microsoft’s diffusion report, and there is a lot we can talk about on whether AI diffusion is leading to equitable socioeconomic benefits and, ultimately, productivity gains in different industries. Part of the problem with these statistics is that they are modelled on formal economies and very industrialised societies, so they are often not relevant to the kinds of economies we might find across the ‘Global South’. I have been part of the AI for Development initiative that the FCDO has been funding for a good number of years. It was initially founded by Canada and Sweden, and I think it has been the largest funding initiative to support AI for development and building AI capabilities across Africa, and with FCDO support it has now moved into Asia as well. It has been an important initiative. With myself and the think tank that we run in South Africa, it supported me in helping with the drafting of the African Union’s continental strategy. It has helped us to engage with the governments of Kenya, South Africa, Zimbabwe, Nigeria, Senegal, Morocco, Egypt and more in supporting local organisations and civil society groups to advance democratic principles, gender equality issues and issues around human rights protection and accountability in these governance frameworks. It has supported AI capacity development in, I think, 13 universities across Africa, which has supported many graduates; a lot of AI innovation for good—for social development, agriculture, health or climate change—has come out of university innovations that this funding initiative has supported. There are so many funding initiatives now, and a lot of them are led by big tech. When I think about what has been successful or not, I ask what the initiatives are that have been led by local priorities. Canada and the UK have been very good at involving local partners and local experts in setting the priorities around what the funding should address, supporting the strengthening of local democratic institutions that are able to sustainably continue this work after funding leaves, and supporting universities that are able to train the next generation of graduates. That is more long-term thinking because a lot of funders want to support exciting innovation that rapidly scales. There have been a number of barriers to innovations scaling to the degree that we might want to see across Africa, and I can speak to that if needed, but we are not necessarily going to see quick wins. Very often the hype and the narrative is around leapfrogging and quick wins, but actually we need to be thinking about long-term institution building, the preservation of democratic values and involving local publics and citizens who are able to contribute and help to shape these debates.
Baroness Crawley
You have answered to a great extent the question that I was going to ask about recent global initiatives. Rachel, I think you feel that the UN Global Dialogue on AI Governance was cumbersome—that is the view that I got from you. What advice would you give those who are now going through the process of the sub-committee aspect of that dialogue in order to make sure that the ‘Global South’ has a real buy-in in the UN dialogue? The second part of my question is this: you talked about 10 African countries having signed up to China’s World Artificial Intelligence Cooperation Organization. Are they countries where China already has a history of trading or infrastructure support?
Dr Rachel Adams
I shall answer the second question first because that will be quicker. I do not know enough about who those 10 countries are and I need to check them, so I do not know enough to be able to comment properly on that. For the UN process, “cumbersome” is probably a good word. What we want to see, and this is where the UK has an opportunity, is thinking about how this credible and legitimate global process can be used to move towards actual concrete decisions, resource allocations where they are needed and accountability. How do we link up that UN process with regional organisations like the African Union or sub-regional organisations like ECOWAS, with regulators, data protection authorities and national human rights institutions that are already regionally and internationally well linked up to the UN, scientific institutions, research centres and funding mechanisms? From my position where I sit—I have been involved in a number of these different processes, from the global summits to the OECD, from UNESCO to the UN—there are now many different multilateral arrangements that are trying to work effectively in this area, and we need to establish some kind of mechanism that links them together so that, for example, we can see what commitments were made at a global summit and whether they were echoed again in the next one. Is there a mechanism to track how far all these big commitments made by different companies and Governments are resulting in the flow of finances to the places that they are talking about? Are people getting trained, in terms of the numbers being cited? We need some kind of mechanism that makes more concrete these very grand arrangements that at the moment are still quite rhetorical and a talking shop. The G20 is an enormous opportunity to build commitments with partners that can be carried into more inclusive institutions. It is an important set of countries being brought together that are able to carry out the commitments that are made in those fora—in the African Union or the EU, and in these other spaces. We can talk more about this, but the UK has a very big opportunity in diplomacy, and the G20 is an important space for moving some of that work forward.
Lord Bruce of Bennachie
Thank you for what you have been saying to us, which I have found extremely interesting, particularly about your own initiatives in Africa, and indeed we have a summary of the African Union continental AI strategy on our papers, which is very clear. You have hinted at my question—you are saying that the FCDO is funding some of these resources—but what is the scope for the UK to take the initiative in engaging with the ‘Global South’ that would be meaningful? You made a point about China and WAICO and the different perceptions of the ‘Global South’ compared with the competition between America and China, but is there space for a country like the UK to engage in a way that addresses the needs of those countries on their own terms? China has said it wants to use AI for development to show people how to do it, but possibly in a different way from us. Is the UK credible if we try to do this? How would we do it? What resources would be required, given that we have pretty well decimated our aid budget? It seems to me it does not have to cost a lot of money but you might think differently.
Dr Rachel Adams
The UK is already doing a lot. The AI for Development initiative that has been running for over six years on the African continent is testament to the work of the UK but also, importantly, the co-operation with Canada, Sweden and now Switzerland, which are also party to it. Perhaps because I spent a lot of time working in institutions across the ‘Global South’, I think that where we need to be putting effort is in supporting and strengthening the capacity of democratic institutions across the ‘Global South’ in places like South Africa, Kenya, and Nigeria, particularly where they have institutional credibility, such as competition commissions that are investigating monopolies; public protection ombudsmen that represent the voice of people where AI is being used in the delivery of public services in ways that might be less than accountable; human rights institutions; and gender equality commissions. It is important to support those organisations to do their work, and particularly for them to engage in these international processes. Often it is the well-funded countries that are able to offer rigorous, robust, multidimensional engagement in these international fora; they are able to send teams of experts to represent them in all these different rooms. There is far less funding in ‘Global South’ countries to participate to that degree. There are experts sitting in human rights commissions, data protection officers and now energy regulators that would be valuable if they had the opportunity to represent their countries’ interests and constraints in these international fora. Importantly, the UK needs to move forward on AI diplomacy. There is a huge gap there, not so much in terms of the need to develop international frameworks, but we need to be training the people and equipping the institutions that are already on the front line of international engagement around AI. The UK has global diplomatic muscle. It is known for its ability to be able to co-operate and deliberate on very thorny issues, where you are balancing very different interests and needs of different countries. It has done that with the nuclear non-proliferation treaty, and it is going to need to think about how it does that again. The UK should use its incredible higher education institutions and its institutional capacity, which is quite neutral, world-class and carries an international brand, in order to provide training to diplomats in ‘Global South’ countries and support the building of relationships between individuals working in different parts of the world on these issues so that, from a slightly more bottom-up perspective, we can build a shared language around what the risks are. We are building a new programme on AI diplomacy at the University of Cambridge. The FCDO is involved in that. We want, and the UK needs, to be an anchor nation on this programme. It draws together all the international links that we have with Governments and international organisations around the world to really centre the UK and the University of Cambridge as a place that can offer training on these issues in a very prescient way.
Lord Bruce of Bennachie
Given what you have told us, it is perfectly reasonable to promote the University of Cambridge’s role in this. However, you mentioned Kenya, for example, where our aid budget next year will be £5 million—almost nothing—and we are currently still in the process of reducing our diplomatic service by something like 30%. So how do we do that? It seems to me everything you say is terrific but the money and the people are not there.
Dr Rachel Adams
We already have the Chevening scholarships, which would fund people from Kenya to come to the UK and participate in a programme like that. Training and education is incredible soft power. It is not as expensive as building data centres or compute infrastructure. The University of Cambridge already has international education standards that are adopted by the majority of countries around the world, and it is time that we deliberately and decisively saw that as diplomatic power that can be used now.
The Chair
You will have heard the previous witnesses answer this question: what would be your top recommendation from this committee to the British Government?
Dr Rachel Adams
Using the G20 opportunity to address and resurface some of the shared commitments that were identified during South Africa’s G20. We have not spoken about this, but the UK has done a lot of leading work in issues that do not seem immediately strategic, like tech-facilitated gender-based violence, and these issues were really pertinent during South Africa’s G20 presidency. I would suggest resurfacing issues like that. The UK being an anchor nation for a global AI diplomacy programme is another important recommendation that feels to me like low-hanging fruit.
The Chair
Thank you, Dr Adams. We are grateful to you for coming along today and being the sole representative at this particular evidence session. Thank you for the information you have given us and the views you have expressed.