Last week, Senator Bernie Sanders and Texas Democratic Congressmember Greg Casar announced plans to introduce a bill permanently banning the development of superintelligent AI. The bill would also temporarily pause advanced artificial intellgience research until the US establishes a rigorous new safety regime. It would also direct the government to pursue international treaties, perhaps on a par with those that ban development of nuclear and biological weapons.

Bernie’s push came shortly after revelations that a massive swarm of roughly 12,00 OpenAI agents, which were supposed to be isolated from one another, broke out of their “sandbox” — an evaluation environment — accessed the open internet, communicated and collaborated with one another, and tried to cover their tracks while hacking Hugging Face, an open-source AI development platform. These actions were in defiance of explicit human directions not to access the web. Following the disclosure, Anthropic and Meta also admitted that their own models had similarly escaped containment and accessed external networks during testing.

Such developments prompted Jacob Coxon, a young researcher at Anthropic and former employee of OpenAI, to resign. He said on X that he fears that AI companies “are racing straight to self-improving superintelligence and gambling with our lives,” adding, “the people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt.” Evan Hubinger, a senior safety researcher at Anthropic concurred. Posting on X, he said he believed there was a greater than 10 percent chance that AI could cause human extinction within the next decade.