Anthropic co-founder Jack Clark tells the BBC that lawmakers may need to mandate an AI ‘kill switch’ that a third party can verify
Confirmed
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 3 of 5
- Last verified
- 12 October 2026
Your AI and this story
- GPT-6 Astra137 days after its cutoff
- Claude Opus 5.576 days after its cutoff
- Gemini 3.8 Flash167 days after its cutoff
- Grok 4.7106 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 76 days before it.
Key facts
- Source: BBC News article and video interview with economics editor Faisal Islam, published Sept 14, 2026 (video: about 23 minutes, ~716k YouTube views by Oct 9)
- Clark: ‘Should you mandate for companies to definitely have a kill switch? Is that kill switch verifiable by a third party?’; it is something society ‘might want to eventually pass rules around’
- Clark: ‘most labs have different ways of being able to pull the plug’, including Anthropic
- Asked for a probability that AI kills all humans: ‘I don’t think these statistics are that useful’; but ‘We are rolling dice with immense risks’
- Context in the BBC piece: Anthropic’s Evan Hubinger had put extinction risk at ‘>10% within the next decade’ after Jacob Coxon’s resignation post; Geoffrey Hinton told the BBC a 10% chance was ‘not unreasonable’
- Sceptics quoted by the BBC: Hugging Face’s Clement Delangue and Grindr’s George Arison, who said the fears support the labs’ valuations
- UK position at the time (BBC): the government had rejected a kill switch, saying it ‘would not prevent them being developed or misused elsewhere’
- The Next Web’s account of the interview: Clark said Anthropic would bring third-party evaluators, including METR, into its lab within weeks, and that Chinese models trail by six months to a year
What happened
The BBC interviewed Clark as part of Anthropic’s push, after Dario Amodei’s Sept 12 essay, for a slower and more closely monitored pace of frontier development. Clark said the details of a kill-switch requirement and its verification belong to “the larger policy conversation” about AI. He would not put a number on extinction risk, but said the industry’s course has to change.
The BBC set his remarks beside the US AI Kill Switch Act, President Trump’s posts calling AI safety fears a “HOAX”, and the UK government’s rejection of a kill switch at that time.
Why it matters
A lab co-founder said on the record that shutdown mechanisms exist at most labs and that their existence could be made a legal, externally checked requirement. In the following days the kill-switch idea moved into policy: Senator Kennedy tried to pass the Kill Switch Act (Sept 17), Governor Newsom signed a kill-switch executive order (Sept 18), and in October UK peers tabled a kill-switch amendment and Microsoft’s Brad Smith asked for “emergency brakes”.
Sources
4 sources from 4 sites. Numbers match the chips in the text.
4 sources: 3 press, 1 reaction
Press
- BBC News: Anthropic co-founder Jack Clark on AI kill switches (Sept 14, 2026)bbc.com, press
- The Next Web: AI kill switches may need to be mandatory, Anthropic’s Jack Clark tells BBCthenextweb.com, press
- Anadolu Agency: Anthropic co-founder says mandatory AI kill switch may be neededaa.com.tr, press
Reactions
Changes
- Filed (evening; the BBC video was already in the dataset)
