Anthropic founder says ‘AI kill switch’ may be needed as Trump declares 'hoax'

Anthropic founder says ‘AI kill switch’ may be needed as Trump declares 'hoax' Jack Clark was speaking after the ‘Godfather of AI’ Nobel prize winner Geoffrey Hinton said there is a ‘10 percent chance’ AI could ‘kill all humans’ in the next decade AI may need a ‘kill switch’ to stop it going haywire, a co-founder of Anthropic, the firm behind the Claude chatbot, has said. Jack Clark, one of the seven co-founders of Anthropic had his say on artificial intelligence safety on Monday. Debate on the technology's development is raging after multiple warnings, including from the ‘Godfather of AI’ Nobel prize winner Geoffrey Hinton who said there is a ‘10 percent chance’ AI could ‘kill all humans’ in the next decade. Hinton's comments came after Jacob Coxon resigned from Anthropic employee over his fears the technology could end civilisation as we know it by 2036. That led to the bosses of the three major American AI firms — OpenAI’s Sam Altman, Anthropic’s Dario Amodei, and Grok’s Elon Musk — to agree external ‘reviewers’ should be installed at ‘frontier’ AI firms. Now, another leading light has told the BBC a ‘kill switch’ might be necessary. "Should you mandate for companies to definitely have a kill switch? Is that kill switch verifiable by a third party?" Jack Clark asked. He continued: "I think that's the kind of thing society is going to want to know and might want to eventually pass rules around." But when pressed on if he thought artificial intelligence had a chance of killing humanity, he said he didn’t ‘think these statistics are that useful’, although he added allowing AI to continue ‘totally unregulated’ was a bad idea. Politicians in the US have proposed the Kill Switch Act which would require companies to have a way to shut down problematic AI tools. However, President Donald Trump has said he believes claims AI could take over the world are a ‘hoax’, potentially stopping the Kill Switch Act in its tracks. In the UK, the government rejected the idea of introducing a kill switch for British-developed AI models only three days ago.