Leadership1 publisher3 min readPublished
Two cabinet ministers set out different UK positions on AI risk on the same day
Louise Haigh will tell the TUC conference that ministers must heed the warnings of AI's own developers, while on the same day the business secretary, Jonathan Reynolds, said talk of a kill switch has little practical meaning.
The Board Room · Leadership desk
What happened
- Anthropic urged its competitors and governments over the weekend to coordinate a global slowdown in AI development, and rivals including OpenAI, Google DeepMind and Elon Musk's X backed the call.
- Jonathan Reynolds, the business secretary, told BBC Radio 4 that discussing a kill switch would not be particularly helpful and that it had little practical meaning.
- The parliamentary committee for human rights called on Monday for a UK regulatory framework with an independent oversight body and legislation to protect the public.
Compiled by The Board RoomSomething wrong?How this is made
Why it matters
- contradiction Two cabinet ministers set out different positions on the same risk on one day, so how a company judges UK timing depends on which department it is dealing with.
- constraint Reynolds tied any UK rule to keeping access to frontier developments. That limits what ministers can sign up to before international partners move.
- decision With OpenAI asking for legislation, anyone lobbying against UK AI rules is now arguing against a frontier developer as well as a select committee.
Haigh's text commits the government to attention. She will say ministers must "heed the warnings of those who are at the forefront of developing this technology" and "be ready to work with international partners" to prioritise public safety and national security [3]. The same speech sets out priorities for public control of transport, water and energy [4]. The speech does not promise legislation. On the benefits she is plain: "AI has enormous potential to transform our public services, make our businesses stronger and deliver new scientific breakthroughs," she will say [5].
Jonathan Reynolds set a limit the same day [16]. He told BBC Radio 4's Today programme: "If you regulate it in a way where you're no longer having access to those frontier developments, that would obviously make us less safe. So I think you've got to be proportionate and understanding about this in how we seek to regulate going forward." [9] He also said: "I think we've got to be careful not to get hyperbolic about this either." [8] His condition is that any UK rule has to leave the UK's access to frontier models intact.
The mandatory kill switch is Anthropic's suggestion. Jack Clark, one of the company's founders, said on Monday night that a switch held by a third party may need to be mandatory for companies, and called it something society "might want to eventually pass rules around" [6]. Reynolds said he did not think it would be particularly helpful to discuss one, and that it had little practical meaning [10].
The pressure that reached the TUC platform started with lab staff. Three Anthropic researchers warned that AI could wipe out humanity within the decade, and Labour MPs and peers then pressed for more cooperation with international partners on regulation [13]. Bilal Chughtai, a safety researcher who left Google DeepMind this summer, said late on Monday that "AI has the potential to kill us all" [15]. Parliament has moved ahead of government here: the parliamentary committee for human rights called on Monday for a regulatory framework with an independent oversight body and legislation to protect the public [11].
OpenAI's position deserves separate attention. The company was among the rivals backing Anthropic's call for a coordinated global slowdown [1], and it has also urged the government to capitalise on renewed safety fears and impose legislation reining in the technology [7]. In my view, a slowdown agreed among the firms already at the frontier holds the current order in place, and rules written to that order cost most to whoever has not built a frontier model yet.
None of this is law. The one future date in the account is next month's evidence session, which Liam Byrne, the Labour chair of the business and trade committee, requested while citing "widely shared concerns about the adequacy of current AI safety governance" [12][17]. The body being asked to answer for that adequacy is the government-backed AI Security Institute, not the labs [12].
What to watch
- Whether Downing Street reconciles Haigh's 'heed the warnings' line with Reynolds's proportionality test, or lets both stand.
- What the AI Security Institute says at next month's business and trade committee hearing about the adequacy of current safety governance.
- Whether the international partners Haigh invokes are ever named, and in which forum a coordinated slowdown would be agreed.