Quote · All-In with Chamath, Jason, Sacks & Friedberg
World's First Trillionaire, Anthropic Fable Banned, The New Oligarchs, Iran Peace Deal
Where this was said
Behind the scenes of Anthropic's Fable ban
At 1:07:54 · chapter starts 33:34
Sacks gives a firsthand White House account: Anthropic expanded Mythos to 50+ companies without approval, Amazon found a jailbreak, Dario refused to take it down, and the export control letter followed.
Anthropic expanded the Mythos preview to 50+ companies without White House approval, including one believed to have China ties. Amazon found a jailbreak in Fable. Dario said no to taking it down when the Treasury Secretary called personally. That's why the export control letter was sent.
Dario came to DC in April and told officials Mythos was a cyberweapon. Mission accomplished — everyone was scared. Then when Fable had a jailbreak, those same officials treated it exactly as Dario had primed them to: as a national security emergency.
Anthropic expanded its Mythos preview program to more than 50 companies without consulting the White House, according to The Washington Post.
Anthropic's mishandling gives Amazon, Microsoft, and Google the perfect justification to become the world's AI gatekeepers. They'll wrap every model in KYC, charge a toll, and call it safety. The frontier labs handed them this on a silver platter.
IBM once owned chips, hardware, OS, and software. The government forced disaggregation and the software industry was born. The same will happen in AI. The current 3-4 player oligopoly won't hold — market forces will fragment every layer of the stack.
Claude's analysis of Dario Amodei found a pattern of 'epistemic exceptionalism' — a framework that always concludes only Dario can be trusted with AI. When your safety model outputs 'me' no matter the input, that's the tell.
