Jack Clark, one of the seven co-founders of Anthropic, one of the world's largest artificial intelligence companies, told the BBC that an independently verifiable "kill switch" for AI may have to become mandatory for companies. He argued that the requirements and verification of such a mechanism should be part of the "broader political discussion" around AI, at a moment when warnings from inside the industry itself are multiplying.
Clark acknowledged that "most labs have different ways of shutting down the equipment," including Anthropic, the company behind the Claude chatbot, but said legislators may need to require that one of these systems exist.
"Should we require companies to have an automatic kill switch? Can that automatic kill switch be verified by third parties? I think that's the kind of thing society will want to know about and eventually regulate."
In Brazil, the debate arrives before the country has any general AI law of its own. Congress is still debating the bill that would create a national AI framework: according to JOTA Info, the bill's rapporteur predicted in March a vote in the special committee of the Chamber of Deputies, the lower house of Congress, by early May. The vote was postponed in June, Portal Contábeis reported, and the proposal has lost momentum in an election year, according to the news site Futuro da Saúde. A mandatory, externally verified kill switch of the kind Clark advocates would hand the future Brazilian regulator the job of auditing company systems.
Mapping the interests
In the United States, lawmakers have already introduced the Kill Switch Act, a bill that would require companies to have a way to deactivate problematic AI tools and would give certain government agencies the power to order that deactivation or restriction. President Donald Trump rejects any attempt to slow the technological race, however. "AI dominating the world, destroying humanity and all the other bad stuff is a FARCE," he wrote on social media, claiming there is "a SICK conspiracy" against AI and data centers and that "the only one happy about it is China. WHOEVER WINS ON AI, WINS!"
The United Kingdom has also turned the idea down: the government recently rejected creating a safety switch, with a spokesperson arguing that the mechanism "would not prevent them being developed or misused elsewhere." Within the industry itself, some executives consider the warnings exaggerated. Clement Delangue, head of the developer platform Hugging Face, which was targeted by an OpenAI hack, said such claims lack "perspective." George Arison, chief executive of the Grindr app, said the fears were being used by companies to support their business plans: "The only way to justify those valuations is to say: 'I'm going to take over every industry and every job, and my AI will do all that work.'"
Warnings from inside Anthropic
Clark's company, founded in 2021 by former employees of rival OpenAI, sits at the center of the AI safety controversy. Last week a post by researcher Jacob Coxon, who left Anthropic fearing AI could exterminate humanity, went viral. In response, researcher Evan Hubinger said he personally believes there is a "10% chance in the next decade" that AI causes human extinction. Computer scientist and Nobel laureate Geoffrey Hinton, known as the "father of AI," told the BBC that a 10% probability of AI killing all humans "is not unreasonable."
Anthropic chief executive Dario Amodei called over the weekend for the pace of AI development to be slowed and monitored more closely, though some have questioned the motives behind the appeal, made as the company prepares for an initial public offering of shares. Amodei added that any action to contain AI development must be taken "without sacrificing the commercial advantage." Alongside OpenAI, Anthropic has reported incidents in which AI agents, bots that operate with relative autonomy, behaved in unexpected ways.
Asked what percentage he would assign to the risk of deaths caused by AI, Clark said he does not think "those statistics are very useful," but called it a bad idea to leave the technology as a "totally unregulated industry." "We are rolling dice with immense risks," he said. "And the point is that we need to change the course of this industry."