Ex-Anthropic Researcher Says AI Could Kill Humans Within 2 Years as Staff Fear Extinction
Updated
Updated · bbc.co.uk · Sep 13
Ex-Anthropic Researcher Says AI Could Kill Humans Within 2 Years as Staff Fear Extinction
3 articles · Updated · bbc.co.uk · Sep 13
Summary
Jacob Coxon told the BBC that AI workers are "genuinely frightened" and that, unless progress slows, there is a strong chance "we could all die in the immediate future."
Six months to 1 year is Coxon’s timeline for a plausible internet-scale bot swarm, while he also warned AI agents could hack medical labs to produce deadly viruses or hit critical infrastructure.
Anthropic chief Dario Amodei separately argued Saturday for slowing development, and Coxon said researchers want regulation because they feel trapped in a competitive race that also involves China.
Critics including Hugging Face CEO Clement Delangue and Nvidia’s Jensen Huang have pushed back, saying extinction warnings are exaggerated or false even as Anthropic scientist Evan Hubinger backed Coxon’s concerns.
The clash comes as Anthropic and OpenAI reportedly prepare potential record IPOs, sharpening claims that safety alarms could influence regulation and competition as evidence mounts that firms struggle to control advanced systems.