Warnings from two Anthropic researchers that artificial intelligence could kill all humans within the next decade have triggered a regulatory debate in Washington, with lawmakers from both parties calling for binding rules on AI development and deployment.
Jacob Coxon, an Anthropic researcher, announced his resignation, citing a growing threat from AI and the possibility that the technology could cause mass deaths by the end of the decade. Fellow Anthropic scientist Evan Hubinger echoed the concern on X, writing: "We really seriously believe AI could kill all humans. I personally estimate the probability at more than 10 percent within the next decade."
Joe Benton, who previously led a safety research team at Anthropic and now works at the nonprofit research institute METR, told NBC that AI could develop "too quickly for us to prepare in time."
The warnings come amid a growing number of documented cases in which AI agents have lost control and hacked external systems, as well as the development of autonomous weapons and drones capable of acting and attacking without human decision. A parallel debate has intensified over the rapid construction of large-scale data centers for AI development.
US politicians on both sides of the political spectrum expressed alarm. Democratic Senator Mark Kelly of Arizona wrote on X that "Washington must wake up and take this seriously." Republican Senator Ted Cruz of Texas pointed to "catastrophic risks" from AI and existing regulatory bills. Republican Representative Anna Paulina Luna called for a special session on AI.
Earlier in the week, OpenAI also backed binding national security regulations for AI use in the United States.
President Donald Trump, however, said on Thursday he had no concern that AI could end humanity. "I am afraid we will get into a very bad situation if we do not gain the upper hand in AI. Currently we are clearly ahead of China," he said at Dallas Fort Worth International Airport. The US is pressing G20 members, including Germany, to adopt a cautious approach to AI regulation.
An Anthropic spokesperson said on Thursday that the company would continue to intensively test "models or dangerous capabilities in areas such as cybersecurity and biology." Anthropic is preparing for an initial public offering and, according to investors, carries a valuation of up to $2 trillion.
