It’s hard to float a near-$1 trillion IPO if one of your key products isn’t even on the market. The looming public offering is adding to the pressure on Anthropic, whose staff members flocked to DC this week to triage with US government officials that imposed export controls on one of the company’s most powerful models on Friday.
Anthropic had brushed off the White House’s concerns that guardrails on its Fable 5 model could be circumvented, adding to the string of fights the AI lab has gotten into with the White House. Most of those conflicts have bolstered the company’s safety-first reputation with the engineers and tech workers it’s fighting to recruit, including its decision to push back against the Pentagon’s use of AI for autonomous weapons.
But IPOs are about storytelling, and Anthropic will have to figure out how to balance the “trust us, we’re the good guys” mantra with its ability to get along with the US government.
“The truth is that no AI model has ever avoided being jailbroken,” Semafor’s Reed Albergotti wrote last week. But arguing, as a group of AI experts pleading Anthropic’s case to the White House did, that Fable is “not uniquely good” at finding and exploiting holes in code risks ceding the moral high ground.