Burnham pitches UK as AI safety leader; ministers split on scope

Britain's new Prime Minister Andy Burnham wants to put AI safety 'center stage' at the UK's 2027 G20 Presidency, pitching a 'new global code to capture its benefits whilst being clear-eyed about its risks.' His own AI Minister, Kanishka Narayan, has told audiences the country has 'already effectively banned superintelligence' through copyright law and a shortage of compute, according to Tech Policy Press's Evie Breese. The two claims sit uneasily together, and that is the story.

Narayan's arithmetic is contested on its own terms. He put UK compute capacity at roughly 1.4 gigawatts and gave a target of single-digit gigawatts by the end of the decade. Breese reports official figures giving the UK 1.6 GW of data centre capacity in autumn 2024, with projections to between 3.3 and 6.3 GW by 2030. If the upper end holds, the infrastructure argument against a domestic path to superintelligence weakens on the government's own numbers.

Parliament is already moving ahead of ministers. Labour MP Alex Sobel introduced a private members' bill in September that would criminalise the development of artificial superintelligence, defined as AI capable of acts that 'neutralize, displace, circumvent, subvert, or render ineffective relevant human authorities.' Over 70 MPs and peers have signed on. The Ada Lovelace Institute, meanwhile, has mapped out four regulatory scenarios; officials, Breese reports, lean toward a narrowly scoped national-security framing rather than comprehensive regulation.

A couple of the policy voices we track shared the piece this week, and the pull was less the summit calendar than the gap it exposes. Britain is pitching the world on AI safety leadership before deciding at home whether that leadership means a Sobel-style ban, a Lovelace-style scenario pick, or Narayan's claim that copyright has already done the work.