AI, Parliament and the UK's Healthcare opportunity

I joined a recent event hosted by Lord Mendelsohn marking the first anniversary of the UK Cancer Vaccine AI Scientist & Supercomputing Project to give the closing remarks.  It was a fascinating discussion on the latest developments in AI, genomics and personalised cancer vaccine research.

Thank you Jon. It’s a privilege to close what has been an inspiring day,  I know, and my thanks to Lord Mendelsohn for bringing us together, to Dame Angela McLean and Dan Cole, to the Oxford team under Dr Lennard Lee  for a remarkable account of the year, and to our sovereign leadership speakers. And I want to thank Steve Easton in particular, because the patient perspective is where all of this begins and ends.

It’s where it begins for me, too. My late first wife, Vicky, was a doctor and research endocrinologist at Barts. When she was diagnosed with ovarian cancer, she found that patients and families had almost nowhere to turn for clear, honest information, so she set out to change that. In 1985 she founded the charity CancerBACUP, now part of Macmillan, and  set up a cancer telephone advice service. She died two years later. Vicky believed passionately that science and candour together could transform what a cancer diagnosis means. I think she would have been thrilled by what this project is attempting, and so this occasion means a great deal to me.

I’ve been asked to set the project in its wider context. I’ve spent much of the last decade warning about what can go wrong with AI: bias, opacity, unaccountable decisions. In my book I asked whether the algorithm would be our servant or our master. Projects like this one are the best answer I know to that question. This is AI working as a servant, in the service of patients, of the NHS, and of one of the oldest and cruellest challenges in medicine.

When I chaired the Lords AI Select Committee in 2017–18, we said that the UK’s distinctive advantage lay in combining three things: a world-class research base, the NHS as a single, unique source of health data, and a reputation for trustworthy regulation. Eight years on, this project brings all three together.

So what is Parliament’s role in making sure this opportunity is realised? Let me suggest three things.

First, trust is the licence to operate. We’ve heard from Dr Ivarsson about unlocking health data. The condition for doing so is trust. We have learnt, sometimes painfully, that the public will support the use of their data for research, but only if they are told honestly what is happening, who benefits, and how they are protected. Trusted research environments, meaningful patient engagement and clear lines on commercial access are not obstacles to this work; they are what will allow it to scale. Parliament -and we have a new Health Bill brining in Single Patient records- should hold government to that standard, and projects like this can show what good looks like.

Second, sovereign capability has to mean something. Professor McIntosh-Smith has shown what sovereign compute can do. My point is about the other half of sovereignty: government as a customer. Sovereignty isn’t achieved by announcements; it’s achieved by sustained investment and by the state acting as an intelligent buyer. Along with colleagues I’ve pressed ministers to reform public procurement so that government becomes an anchor customer for British AI innovation, rather than defaulting to the largest overseas providers. Health is where that argument is strongest. If we build world-leading models on NHS data with British science, we should make sure the value, the know-how and the control stay here for the benefit of patients.

Third, good regulation is an accelerator, not a brake. AI scientists and autonomous labs raise new questions. Who is accountable when an AI system proposes a therapeutic target? How do we validate models whose reasoning we can’t fully inspect? How do regulators such as the MHRA keep pace? My consistent view is that we need clear, proportionate and, for the highest-risk uses, mandatory standards, with a principle of human-in-command: humans setting the goals, scrutinising the outputs and taking responsibility for the decisions. The HFEA showed that firm, intelligent regulation can make the UK the place where frontier science is done. We should aim for the same in AI and medicine.

Underpinning all of this is a simple point about time. Vaccine development can run over many years; political and spending cycles tend not to. One of the most useful things parliamentarians can do, across all parties, is to insist on long-term, stable commitment to compute, to skills, and to the clinical research infrastructure that lets work like this move from lab to bedside.

Let me end where I began. Vicky’s conviction was that patients deserve both the best science and the whole truth. That is a good test for AI in medicine  too. The question is never whether AI is powerful; it is what we choose to use that power for, and on whose terms. Here we have a cause to champion: AI as a servant of science and of patients, built in Britain, on foundations of trust. Parliament’s job is to make sure the conditions are there for it to succeed, and I know colleagues here, share that commitment.

Happy first birthday to the project. I look forward very much to the second. And now, I believe, there is cake!

Thank you.