AI 'kill switch' may need to be mandatory, Anthropic co-founder says
BBC News ● Covered by 27 sources
Anthropic’s Jack Clark says AI firms may need a mandatory kill switch checked by outsiders. He says the industry is gambling with huge risks, and regulators may have to step in.
Based on reporting by BBC News — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Anthropic co-founder Jack Clark says companies may eventually need to be forced to build a way to shut down AI systems completely, and that the shutdown mechanism should be verifiable by a third party. He told the BBC this is the sort of thing society may want rules around, even if many labs already have their own way to pull the plug.
His comments land in the middle of a louder-than-usual fight over AI safety. In recent days, Anthropic chief executive Dario Amodei called for AI development to slow and come under tighter monitoring, while critics have questioned what sits behind that message. Clark did not offer a percentage for the chance of AI wiping out humanity, saying those numbers are not very useful. But he was blunt about the direction of travel: letting AI remain a totally unregulated industry is a bad idea.
That warning is not coming out of nowhere. Last week, a former Anthropic researcher went viral after quitting over fears that AI could wipe out humanity. Anthropic scientist Evan Hubinger then said he personally thought human extinction from AI was more than 10% likely within the next decade. Geoffrey Hinton, the Nobel Prize winner often called the Godfather of AI, told the BBC a 10% chance of AI killing all humans was not unreasonable.
The policy split is already visible. US lawmakers have proposed a Kill Switch Act that would require companies to be able to shut down problematic AI tools, and would let some government agencies order a tool turned off or limited. Donald Trump has dismissed warnings about AI taking over the world as a hoax, while the UK government has rejected a kill switch on the grounds that it would not stop the tools being built or misused elsewhere. Anthropic itself, formed in 2021 by former OpenAI employees, has released a string of more capable models this year and is now reportedly preparing for a potentially record-setting IPO.
My take — AI-written commentary, not fact-checked reporting
This is the part where the industry pretends safety is a branding issue until lawmakers make it one. A mandatory, checkable shutdown mechanism is not sci-fi; it is the least dramatic response to a technology whose own builders keep describing as dangerous. If companies want the freedom to ship ever more powerful systems, they should expect the leash to get shorter, not longer.
Read more about this at: BBC News
