Multiple current and former Anthropic employees warned — and admitted — on Tuesday that the researchers building new AI models believe the technology could lead to humanity’s destruction by the end of the decade, reflecting the industry’s frenzy over reining in rapid development.
Jacob Coxon, a former AI researcher at Anthropic, [wrote](https://x.com/hilbertspaess/status/2097476196791709843?s=20) on X that he resigned from the company on Tuesday because he believed the Dario Amodei-led company and [OpenAI](https://variety.com/t/openai/), his former employer, were “gambling with our lives” in how they improved the technology.
“The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon wrote. “This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible – but I hear the same people express fear privately. No other human activity poses this level of danger.”
Popular on Variety
Anthropic did not respond to an immediate request for comment, but Coxon’s warnings were surprisingly affirmed by Evan Hubinger, Anthropic’s alignment scene lead, and Samuel Marks, who runs the company‘s scalable oversight division.
Hubinger wrote the company did “earnestly believe AI could kill all humans” and that he thought it was a greater than 10% risk “within the next desk.”
“I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to,” he [wrote](https://x.com/EvanHub/status/2097497037956891126?s=20) on X.
Marks’ own X [post](https://x.com/saprmarks/status/2097570226804011302), which he said he wrote in his personal capacity, said that AI researchers continue building the technology despite the risk of “human extinction” because of a financial incentive and the fear the technology would be developed by AI developers who may abuse its capabilities.
“Many AI developer staff desperately want to slow down to figure out how to build AI more safely,” he wrote. “I work on safety research at Anthropic because I hope my work will reduce the chance of these extinction-level bad outcomes.”
The comments from Anthropic employees reflect a broader public tonal shift among AI companies, many of which entered the market touting the transformative potential of AI models but now warn that those models risk escaping human control. Anthropic as a company has [called](https://www.anthropic.com/institute/recursive-self-improvement) for a development slowdown since June, while OpenAI paired its debut of GPT-6 Astra last week, which it claimed pushed the world into an era of artificial general intelligence, with several warnings by company leaders and researchers that a slowdown was needed to improve safety conditions.
“No one is prepared for the consequences of a continued rapid rise in machine intelligence,” wrote Jakub Pachocki, OpenAI’s chief scientist, in a [blog post](https://openai.com/index/an-alien-mind/) on Sunday that was later reposted by OpenAI CEO Sam Altman on X.
Still, the industry sentiment has received pushback from the Trump administration, which has argued that slowing down development risks ceding the AI arms race to a rapidly evolving China.
“There is no day after tomorrow if China wins at this,” Treasury Secretary Scott Bessent [told](https://www.breitbart.com/politics/2026/09/08/exclusive-secretary-scott-bessent-if-china-were-pull-ahead-of-us-on-ai-nothing-else-matters/) Breitbart on Tuesday. “Like everything, a trillion-and-a-half-dollar defense budget, the Iron Dome. If they were to pull ahead of us on AI, then nothing else matters.”