Worried Anthropic researchers warn that AI ‘could kill all humans’ — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Jacob Coxon resigned from Anthropic, posting on X that he quit over the company's "lax approach to safety" and accusing both Anthropic and OpenAI of "racing straight to self-improving superintelligence and gambling with our lives."
- Evan Hubinger, who leads one of Anthropic's AI safety teams, said he personally estimates a greater than 10% chance AI could kill all humans "within the next decade" and warned that self-improving AI "is happening faster than we thought."
- Hubinger conceded that Anthropic "does not yet have a plan" for ensuring advanced AI remains safe and aligned with human values and is "not clearly on track to" develop one either.
- Coxon previously trained systems for OpenAI; his departure is one of the most high-profile safety-driven exits from Anthropic, a company founded by former OpenAI members who themselves left over safety concerns.
- The exchange comes as AI companies prepare for anticipated IPOs and work through the fallout from "numerous rogue agent incidents" and high-profile warnings about the monitorability of frontier models.
Why it matters: Anthropic's IPO narrative just got more complicated: a sitting safety team leader is publicly conceding there is no plan to keep superintelligence safe while putting extinction odds above 10%. With a high-profile internal resignation and a second insider endorsing the warning on the record, existential-risk disclosures shift from internal debate to public investment calculus — one that prospective public-market investors and outside watchdogs can now cite directly.
Ask SkimNews
