Investor and executive scepticism at a high-profile San Francisco conference threatens to blunt momentum for urgent action on artificial intelligence. Between pitches about growth and returns, leaders in technology and finance publicly questioned stark claims from current and former Anthropic staff that AI could pose existential danger.
The debate was sparked by the abrupt resignation of Anthropic researcher Jacob Coxon, who wrote that people building AI were "gambling with our lives" and warned "these will soon be superhuman systems that can hack anything." Other insiders amplified the alarm: a team lead at Anthropic, Evan Hubinger, posted that "We really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade."
Some executives and investors responded with scepticism, suggesting the warnings may overstate risk or even serve other aims. Several speakers at the conference said the comments were untrue or excessive. Grindr chief executive George Arison called the tone of those statements "dangerous" and said he told engineers to stop using Anthropic's technology. Nvidia chief executive Jensen Huang dismissed the idea that AI "is going to be the end of humanity" as "complete nonsense".
Critics also pointed to larger incentives. Anthropic was valued at $965bn (£713bn) in its most recent fundraising round, and both Anthropic and OpenAI are reportedly preparing for potentially record-setting initial public offerings. Some suggested dramatic warnings could bolster perceptions of product power or encourage regulatory settings that reshape competition.
Anthropic has itself flagged technical concerns. The company said its Mythos system could outperform humans at certain hacking and cybersecurity tasks and that it had detected and disrupted threat actors attempting to use its technology for activities including the development of bioweapons and cyber-espionage. Anthropic's chief executive, Dario Amodei, has called for a slowdown in model development and for global regulation, language that several industry figures later described as a constructive step.
The exchange at the conference underscored a widening divide inside the AI sector: some insiders pressing urgent safety warnings, others questioning their accuracy or motives. The immediate consequence is a more fractious debate over how quickly to regulate and which industry actors should shape those rules. The next phase will hinge on whether regulators, investors or firms themselves translate these competing positions into concrete policy or market responses.
