Six months after the launch of the Government’s AI Opportunities Action Plan, artificial intelligence trade body UKAI, hosted a significant review event to assess the plan’s progress.
The review examined delivery of all 50 of the Action Plan’s recommendations, all of which were accepted by the Government in their official response to the Plan.
The second of a series of panels was titled: “Getting the Foundations Right”. Chaired by former AI Minister Matt Warman, the discussed aimed to evaluate the foundations that the success of the AI industry in Britain rests upon.
Joining Matt Warman on the panel were Gaia Marcus (Director of the Ada Lovelace Institute), Fiona Ghosh (Technology Law Partner at Ashurst LLP), Amir Malik (Alvarez & Marsal’s AI and Digital Transformation Lead), and Sophia Ignatidou (Group Manager for AI Policy at the Information Commissioner’s Office).
The AI Opportunities Action Plan: Answering the Right Problems?
The conversation opened with one fundamental question: What is the single foundational barrier to innovators and stakeholders in AI – Gaia Marcus kicked off an initial response. She said: “Two reflections. First: you can’t implement 50 recommendations. What I haven’t seen yet is a proper analysis of the UK’s comparative advantage, where should we really be focusing? Most UK businesses are going to be at the cheap adoption end of AI. We see big gaps around governance and liability: there’s still no real clarity on legal regimes.
She continued: “Second: we’re not seeing the “big bangs” that drive adoption. Fundamentally, for us, it comes down to people – and the question we still don’t have an answer to is: What are we doing in this country that reflects the attitudes and expectations of our population?”
Those remarks framed the rest of the discussion, which repeatedly returned to the need for a clearer, more focused strategic vision. Fiona Ghosh of Ashurst LLP echoed the concern, noting that clients consistently ask: “What is the global standard?” and “What do we comply with?”
She pointed out that while the UK doesn’t need to be “uniquely British” in its regulation, what it must offer is openness and, critically – “mature frameworks” that give businesses confidence to operate. She warned that without such frameworks, the complexity of overlapping initiatives could stall growth and innovation.
Sophia Ignatidou of the ICO added further insight to Fiona Ghosh’s comments on regulations, explaining that the UK still faces major blockers: lack of transparency around training data, limited governance tools for those procuring AI, and uncertainty over the legal standing of foundational models.
Infrastructure: The Beating Heart of AI Ecosystems
Next, the panel moved on to the topic of infrastructure. Amir Malik set out his view that infrastructure is a critical component of a successful AI ecosystem. He urged that Government should take an active role in creating, maintaining and leveraging infrastructure to forge successful AI in the UK: “On infrastructure: governments need to start acting like platforms. Look at the UAE and Switzerland – they’re already building ethical LLMs. And think about companies like Google, they own their entire infrastructure stack.”
“Governments need to think about sovereign infrastructure the same way. For example, Google laid its own transatlantic cable. Sovereignty isn’t just about compute: it’s about data transfer and infrastructure control.”
Fiona Ghosh echoed his comments, reiterating the unique importance of infrastructure to flourish AI: “Completely agree. We’re [Ashurst LLP] involved in the subsea cable space, GPU and TPU infrastructure, AI factories, there’s real business momentum.”
What’s Really Holding Back Adoption: Is it Public Trust or Product Maturity?
Following the infrastructure discussion, Matt Warman asked whether the real barrier was public trust or product maturity. Gaia Marcus drew a distinction: “We need to separate trust from trustworthiness. Trust isn’t always a good thing in itself, trustworthiness is.”
She cited a recent Ada–Turing survey which showed that 72 per cent of people want AI to be regulated: up ten per cent in two years. While there is growing awareness of AI’s potential, Marcus warned that public concern around scams and deepfakes is rising faster than confidence in the technology.
She also cautioned that those in policymaking rooms are not representative of the wider population. She said: “In the public sector, where expectations are higher, people are sceptical of US tech firms in local services, and they want transparency and redress.” She pointed out that risk is being “pushed downstream,” with businesses potentially liable for problems inherited from upstream foundation models.
Sophia Ignatidou cited research conducted by the ICO, that backed up Marcus’ comments – she stated that consumers are aware of the risks of AI, but that the legal tools available to regulators need to be paired with greater resources and speed – to enable timelier delivery.
Six Months In: Final Thoughts
As the panel drew to a close, Matt Warman asked each panellist for a final reflection: how optimistic were they about the UK’s trajectory, six months into the AI Opportunities Action Plan?
Gaia Marcus responded: “We’re in a tight geopolitical spot. The tech is moving faster than our ability to govern it. The duality with the US, where AI is framed as a national security issue, makes real governance difficult. Our government seems more gung-ho than the public wants it to be. And because proper safeguards aren’t in place, the risks fall on the most vulnerable.”
Fiona Ghosh commented that, six months on from the publication of the AI Opportunities Action Plan, she is more optimistic than she once was: “Yes, we still have major productivity issues, and a huge financial hole to dig ourselves out of. But without ambition, we won’t get there. We need to set a direction of travel, and we’re finally beginning to do that.”
Amir Malik closed with a cautiously hopeful message: “I’m generally optimistic: but there are real risks. From a business and innovation perspective, the UK has a lot going for it. We’ve got world-leading researchers, frontier startups, a strong AI ecosystem in London, UCL, Cambridge, etc. The talent is here.”
“But there’s no escaping it: AI is learning from us: our WhatsApps, our images, our messages. We don’t even know what’s being used. We need stronger data sovereignty, and real user control. We’ve trained LLMs on author content, and even the authors couldn’t tell the difference. That’s how advanced this is getting. And unless we build trust, the risks to privacy and misinformation will only grow.”
“Still, the opportunities in healthcare, assistive tech, robotics – those are real. I think the UK has a shot at leading in this space. But the pace needs to pick up.”
