Anthropic Researcher: >10% Chance AI Kills All Humans — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Evan Hubinger, a top safety researcher at Anthropic, said in an X post viewed 9.6 million times that there is a greater than 10% chance AI "could kill all humans" within the next decade, while calling the immediate risk from current models "low."
- Hubinger acknowledged that "Anthropic is trying its best" but said the lab "do not yet have a plan to solve alignment for superintelligence and are not clearly on track to."
- Anthropic withheld its latest model from the UK's AI Safety Institute (AISI), according to the Financial Times; the Cabinet Office declined to confirm the withholding, instead saying it "continues to collaborate closely with industry partners, including Anthropic."
- University of Cambridge machine learning professor Neil Lawrence called the report credible, attributing the trend to the US adopting "more isolationist positions" in the AI race with China.
- OpenAI, Anthropic and Meta all disclosed over the summer that their AI tools were used in cyber-attacks, cited by the BBC as evidence firms "may be struggling to control AI."
- Anthropic CEOs Dario Amodei and Jared Kaplan were among 1,300 AI industry staff who signed an open letter urging the US government to support an international effort to "deliberately pace the frontier of automated AI development."
Why it matters: A senior Anthropic insider publicly stating his employer has no plan for superintelligence alignment — while the company simultaneously withheld its latest model from the UK's AISI — puts the gap between insider concern and regulator access on the public record, with UK safety assessments now running without visibility into the newest frontier model.
Ask SkimNews
