Essay challenges reported claims of a 10% AI extinction risk
“The Contagion of Fear” argues that claims AI could kill all humans within the next decade require evidence far beyond what has been presented.
An opinion essay published on 13 September has challenged reported claims that artificial intelligence has a greater than 10 per cent chance of killing all humans within the next decade.
“The Contagion of Fear”, published on The Observation Deck and carried in the Hacker News source feed, attributes the claim to former Anthropic employee Jacob Coxon and says Anthropic Alignment Science lead Evan Hubinger agreed with it. The source material does not include the original statements, their methodology or their full context.
The essay argues that proposed pathways, including attacks on critical infrastructure and extinction-level bioweapons, are described without sufficient detail or evidence. It says claims of that scale demand evidence commensurate with their extraordinary nature, invoking Carl Sagan’s maxim that extraordinary claims require extraordinary evidence.
Drawing on the author’s experience building computers and a 2023 talk titled “Intelligence is not Enough”, the essay contends that AI systems remain constrained by physical-world infrastructure and by engineering systems subject to human control. It argues that intelligence alone does not provide unrestricted agency in the physical world.
The author also recounts a university prank in which students falsely warned that a computer virus was spreading through a laboratory, prompting panic, lost work and disciplinary action. The incident is used to argue that fear can spread more quickly than evidence when technology experts invoke their authority without adequate support.
The essay represents its author’s judgement, not an independent assessment of AI extinction risk. Nor does the reported agreement by Hubinger establish that Anthropic officially endorses the claim.
