TL;DR
A cross-party committee of peers and MPs has issued a 100-page report demanding dedicated AI legislation, arguing that the rules Britain has now leave people unprotected. It wants one statutory regulator instead of the current spread of bodies, graduated duties that scale with risk, and a short list of applications banned altogether. It arrives days after ministers rejected a separate parliamentary attempt to legislate an emergency shutdown power.
What the committee asked for
The Joint Committee on Human Rights, whose twelve members sit across the Commons and the Lords, and include Conservative, Liberal Democrat and Labour parliamentarians, published its findings on Monday. Its chair, the Labour MP Alex Sobel, put the problem in blunt terms: “Nowhere in the world, including the UK, has a current legislative and regulatory approach to AI that is fit for purpose.”
Three asks sit at the centre of the report. The first is a single oversight body established in statute, replacing an arrangement the committee describes as “fragmented and difficult to navigate” and prone to “leaving gaps in protection”. The second is a tiered set of duties, weighted towards higher-risk models and applying across the whole lifecycle rather than at a single approval point. The third is the sharpest: certain applications should be off the table entirely, with subliminal techniques and misused profiling or biometric data named as candidates.
The harms cited are not hypothetical. The committee points to faces being scanned without permission, and to sexualised imagery generated of girls and women.
Ministers are pulling the other way
The government has just refused a narrower proposal. Lord Clement-Jones asked peers on 1 September to create a legal power to shut down a frontier model behaving dangerously, and Sobel raised it in the Commons. The Cabinet Office told the BBC that Britain “cannot simply turn AI off”, noting that restricting a model here does nothing to stop it being built or abused abroad. That does not kill the bill, but it makes passage improbable.
Labour’s 2024 manifesto promised binding rules for the handful of firms building the most capable models, and the King’s Speech that July pledged legislation. This year’s King’s Speech contained no AI bill. Education minister Georgia Gould pointed instead to deepfake measures, cyber security legislation and the AI Security Institute’s model testing.
Looking forward
The committee has given a name and a shape to something Whitehall has spent two years avoiding. Under Andy Burnham’s restructuring, which dissolved the science department and dispersed its functions, there is now no single ministry that would obviously own such a bill. UK firms should read the prohibition list as the clearest signal yet of which biometric and profiling deployments carry political risk.