UK lawmakers give government two months to answer for AI's human-rights gaps

Britain's Joint Committee on Human Rights (JCHR), a cross-party panel of MPs and peers chaired by Labour's Alex Sobel, published a report last month concluding that existing UK law cannot protect people from the harms AI is already causing, and that the government has run out of road on its 2024 promise to legislate. With the government's two-month deadline to respond now ticking down toward mid-November, the report has become the sharpest test yet of whether Westminster will move from voluntary pledges to a statutory regime with enforcement powers — a question that matters directly to any company selling or deploying AI into the UK market, and to investors pricing regulatory risk into that exposure.

What the committee found

The report, titled Human Rights and the Regulation of AI, followed an inquiry that opened in July 2025, drew more than 70 written submissions and ten oral evidence sessions, and was formally agreed by committee members on September 9 before publication on September 14. Its central finding, in Sobel's words, is that AI is 'moving with such speed and complexity that its impact is hard to accurately predict,' leaving Britain — like every other jurisdiction the committee examined — without a legislative and regulatory approach that is 'fit for purpose.'

The committee's diagnosis is structural, not just a complaint about pace. It found that the UK's AI Security Institute, the body meant to test powerful models for safety, has no statutory power to compel access to those models or to block a release it judges too risky; developers engage with it purely on a voluntary basis. Existing statutes that could in theory cover AI harms — the Equality Act 2010 and the Human Rights Act 1998 — mostly bind public authorities or discrimination in services, leaving business-to-business AI developers largely untouched. And automated-decision safeguards recently added to UK data protection law are, in the committee's assessment, undermined in practice because a human nominally 'in the loop' on a decision does not guarantee any meaningful human judgment actually occurred.

Nowhere in the world, including the UK, has a current legislative and regulatory approach to AI that is fit for purpose. — Alex Sobel MP, chair of the Joint Committee on Human Rights

What it wants Parliament to do

The committee's ask is a full AI Bill — the legislation promised in the King's Speech of July 2024 but never introduced — built around a risk-tiered framework with several concrete, enforceable elements: prior approval from an oversight body before the highest-risk AI systems can be deployed; due-diligence obligations running across the entire supply chain, not just the company facing the end user; a statutory duty to tell people when AI materially shaped a decision about them; and outright bans, subject to public consultation, on uses such as subliminal manipulation techniques and inappropriate biometric profiling. It also wants the AI Security Institute converted from a voluntary partner into a statutory regulator, and a published timetable for the UK to ratify the Council of Europe's Framework Convention on AI, which it has signed but not yet ratified — the EU, by contrast, ratified it back in May 2025.

Legal practitioners reading the report see it as evidence that Britain's long-standing preference for sector-by-sector, principles-based AI oversight — enforced through existing regulators like the Information Commissioner's Office and the Equality and Human Rights Commission rather than a dedicated AI law — is coming under real strain. Sophie Sheldon, a partner at Simmons & Simmons, has pointed to the report as reinforcing the sense that the UK's sector regulators are struggling to keep pace with AI's risks to fundamental rights, and suggested the pressure for a single legislative answer, closer to the EU AI Act model, is building. The committee itself heard the opposite case from industry during its evidence sessions: Google argued for continued sector-specific rules and warned against duplicative overlapping laws, while Meta characterized the government's current approach as thoughtful and sensible.

Why decision-makers should watch the clock

The government is not legally bound to accept any of the committee's 20 recommendations or 39 conclusions, and ministers have previously argued existing law is adequate. But the formal two-month deadline for a substantive response — falling around mid-November — forces a public reckoning that the government has largely avoided since its 2024 legislative promise lapsed. The political backdrop has also shifted since July, when a change of prime minister saw the Department for Science, Innovation and Technology abolished and AI policy folded into a new ministerial portfolio under Kanishka Narayan, adding fresh uncertainty about who in government actually owns the decision. For companies building or deploying AI into the UK, the practical stakes are a possible shift from today's lighter-touch, multi-regulator landscape toward something closer to the EU AI Act's pre-market approval and supply-chain due-diligence model — a change that would raise compliance costs and timelines but also, proponents argue, give firms a single clear rulebook instead of navigating several regulators' overlapping and sometimes contradictory expectations.