As Anthropic heads towards a $2 trillion IPO, some of the loudest critics are company insiders

Anthropic’s march toward a reportedly record-breaking $2 trillion IPO is colliding with something that would make any banker wince: several of its own researchers are publicly talking about the chance that the company’s AI could kill everyone.

The loudest examples came from inside the house. Evan Hubinger, who leads Anthropic’s Alignment Science team, said he believes AI could kill all humans and put the odds at more than 10% over the next decade. Another researcher, Drake Thomas, said he would burn his equity “to the ground” for a 1% better chance of surviving the situation. That is not exactly the kind of messaging companies usually want floating around while they’re preparing to sell stock to the public.

Still, this isn’t some sudden conversion to doomposting. Anthropic was founded as an AI safety lab, and CEO Dario Amodei has said his own p(doom) sits somewhere between 10% and 25%. The company’s researchers have been warning about catastrophic AI for years, and in that sense these posts are just the company talking more loudly about what it has always been built around.

The awkward part is the timing. A public listing would put Anthropic under shareholder pressure to grow, which may not leave much room for the sort of voluntary slowdown Amodei has floated. It also raises a basic investor question: if the company’s own people say there is no plan yet for controlling superintelligence risks, how exactly should the market price that?

Some investors are already pressing that point. This week, SOC Investment Group, a labor-affiliated shareholder group, called on Anthropic to delay the offering, saying its confidential filing in June came before recent hacking incidents, extinction warnings, and pacing proposals, so investors have not had the full picture. But there is another school of thought too: that going public could force more transparency and scrutiny, which might actually make a safety-focused company safer. OpenAI, meanwhile, has already stepped back from its own IPO plans, with Sam Altman calling it “an ill-advised moment to go public” because of AI safety concerns. When the S-1s finally arrive, the risk sections may do a lot more talking than the roadshow.