21 May Let’s Talk AI makes it to Parliament!
On 18th May, Susan gave evidence to the All Party Parliamentary Group on AI. Let’s Talk AI was central to this evidence. You can read what Susan said below
Speaker intro:
I am Susan Oman, a Senior Lecturer / Associate Professor in Data, AI and Society at the University of Sheffield. Here I lead on Human-centric AI via its Centre for Machine Intelligence.
My expertise lies in understanding how different parts of society respond to technological progress and policy decisions. To pick up on the theme of public AI mentioned by colleagues, it speaks to questions of what public AI is that is contingent on who and what we mean by the public.
My research spans different populations – whether that is across those we might consider a non-expert citizenry – to professions including national statistics offices and Whitehall, to local government offices, and small businesses.
AI sovereignty emerges as a concern in the research I do and the conversations I have – daily – across these stakeholders.
Sovereign AI Fund and local start-ups
In the past week, the Sovereign AI Fund[i] has featured in these discussions with local SMEs who are concerned about limits to the fund’s potential for transformation and confusions over eligibility. One recipient is reported to be incorporated in the US with most senior team members based in Silicon Valley and some in London.
Since the Opportunities Action Plan[ii], the small businesses and start-up culture of Sheffield and the regions have eagerly awaited an announcement along the lines of Sovereign AI. Sadly, they currently feel alienated by confusion over who counts and what counts in this initiative and struggle to see how it contributes to national resilience or competition.
AI sovereignty should be people-centred
We should pause, then, to think about what we mean when we talk about AI sovereignty and how it translates across publics.
- Who is making what, for what purposes, and why?
- How will this technology be used, and crucially, what does it enable to happen next?
- Central to the question for policy-makers, of course, is always who benefits and where benefits.
For sovereign AI, I argue we are ultimately asking a set of deeply human and societal questions that exceed technical, material and macroeconomic questions
The AI Skills Hub[iii] was an opportunity to see past British AI in these terms and consider the importance of future AI infrastructure as investing in people and skills. Following February’s launch, the hub received much criticism, including an op-ed written by myself and colleagues[iv] outlining our concerns over decisions surrounding the £4.1mil investment from the perspective of sovereignty.
Launch included 595 courses, with only 14 benchmarked by Skills England. All these foundational courses were from US providers. Press releases foregrounded “free” training for all, but 60% require payment, including some marked as free. Educators criticised the poor quality of provision, some of which are 10 years’ old, or no longer exist. Some aren’t relevant to the UK, or worse, misrepresent UK law on key issues to sovereignty, including intellectual property.
Therefore, in not considering sovereignty in the delivery of AI literacy, we have invested 4.1million on infrastructure that is far from accessible to all, and what can be accessed trains the UK citizenry on how to US products.
Let’s Talk AI: people-centred AI infrastructure
Policy-making favours the familiar. But when we do things differently, it works. One response to the failures of the Skills Hub that I have been working on, is a collaborative public AI Awareness campaign called Let’s Talk AI.
We went directly to communities with the lowest digital literacy to ask them what they actually wanted from AI literacy as a process and an infrastructure. I urge anyone interested in digital inclusion to look at the outcomes of that project. This people-first approach to AI literacy can be found on the website LetsTalkAI.Org.uk.
The communities we spoke to have wants for AI literacy that are deeply rooted in what they are concerned about, specifically what they trust, (and lack of confidence around what they should trust). These concerns are related to issues of AI sovereignty in thinking about who is using what; who benefits or could be at risk.
Public Concern about AI is rising
And we know from our work on the Public Voices in AI[vi] research project, an £850k UKRI investment, that concern is rising. It increased by 10% between two waves of our nationally representative survey in 2023 and 2025.[vii]
We know in follow-on research from the Ada Lovelace Institute that the UK public do not trust private companies to self-regulate, with 89% wanting an independent regulator for AI, equipped with enforcement powers.[viii]
We also know that 91% of the public feel it is important that AI systems are developed and used in ways that treat people fairly, and that this should be prioritised over economic gains, speed of innovation and international competition.
When the public is asked specific questions about sovereignty, most recently from commercial research at FSS Global, it seems that the public explicitly supports sovereignty – especially considering geopolitical tensions. This is no surprise. What is interesting is that it seems that the public is willing to pay extra for technology developed by domestic companies – two in five state they would be willing to pay 50% more.[ix]
Sovereignty – a caveat
Sovereignty is not an answer without limitations – and these must also be noted. Research I conducted in COVID and shortly after the introduction of UKGDPR spoke with BBC R&D departments who were trialling experimental models of data sovereignty that aimed to reimagine control over personal data.[x] In short, they involved considering data sovereignty at the personal level.
On our Living With Data project[xi] we found a number of these thought experiments weren’t legible to the publics who were shown the systems. The public struggled to understand the need for such a product – they couldn’t imagine how it would fit into their lives. This was because it felt and looked so alien from the platform UX familiar to them. Namely provided by US Big Tech.
Crucially, then, our data literacy in AI-enabled infrastructure is a barrier to how we might imagine AI sovereignty in productive ways that suit the wants of AI sovereignty. Addressing this requires much care and consideration from multiple perspectives, as well as investment and backing from central government.
Conclusion
Sovereignty is crucial for all considerations around British AI and its future. However, we won’t make the decisions that serve the UK without thinking about AI sovereignty as a profoundly societal question and one that demands multiple perspectives.
If we want a resilient AI ecosystem, we must stop asking what the technology can do for our economy at macro scale without considering the economics of everyday choice and human behaviour.
We are beginning to see indications that AI sovereignty increases public trust and therefore support – including purchasing power, but there are also indications that AI sovereignty is limited by how we already live and work and what we know. For that reason, I propose we need to approach a people-centred approach for AI sovereignty.
Let’s Talk AI is a one model of a people-centred approach to AI infrastructure that encourages healthy conversations around AI. Our evaluations indicate that this people-centred approach to infrastructure – and before it is established – is the best way to intervene in people’s increasing concerns surrounding AI.
These alternative models should not only be informing delivery of infrastructure, but crucially communications around infrastructure decisions – and those key questions of the who, what and where of AI that drive policy decisions around AI sovereignty.
True sovereignty begins when we give the people of this country a say in who makes this technology, how it is used, and crucially, what kind of Britain it enables to happen next.
Thank you.
[i] https://www.sovereignai.gov.uk/
[ii] https://www.gov.uk/government/publications/ai-opportunities-action-plan/ai-opportunities-action-plan
[iv] This articles is authored by Tania Duarte, founder, We and AI; Bruna Martins, director at Tecer Digital; Dr. Elinor Carmi, senior lecturer in data politics and social justice, City St. George’s University of London; Dr. Mark Wong, head of social and urban policy, University of Glasgow; Dr Susan Oman, senior Lecturer, data, AI & society, The University of Sheffield; Ismael Kherroubi Garcia, founder & CEO, Kairoi;Cinzia Pusceddu, senior fellow of the Higher Education Academy, independent researcher; Dylan Orchard, postgraduate researcher, King’s College London; Tim Davies, director of research & practice, Connected by Data; Steph Wright, co-founder & managing director, Our AI Collective: https://www.computerweekly.com/opinion/The-UK-governments-AI-skills-programme-betrays-UK-workers-and-our-digital-sovereignty
[vi] https://digitalgood.net/research/public-voices-in-ai/
[vii] https://digitalgood.net/outputs/public-voices-in-ai-outputs/
[ix] https://fgsglobal.com/radar/tech-radar
[x] https://livingwithdata.org/current-research/interviews-focus-groups/