AI policy
The government switched off the best AI model in the world over a bug it won't show us
Three days. That's how long Fable 5 lasted.
Washington forced Anthropic to kill Fable 5 over an unproven 'jailbreak' — and won't show its work.
Timeline: 9 June, Anthropic launches Fable 5. 12 June, 5:21pm ET, a directive lands. By Friday the model is dark — Fable 5 and the restricted Mythos 5, gone. Not throttled. Off. For everyone, because the order covers every foreign national, including Anthropic's own non-American staff.
The receipts are missing
Anthropic's account is blunt: it got verbal notice of a 'potential narrow, non-universal jailbreak,' watched a demo that turned up a handful of already-known minor bugs, and flatly disagrees it warrants pulling the model. The government's actual reasoning? Not published. We are asked to accept a kill order on a model based on a vulnerability nobody outside the room has seen.
And the kicker
Anthropic's own rebuttal points at OpenAI's GPT-5.5, arguing rivals have the same capability. So either the risk is real and universal — in which case why is only Anthropic dark? — or it's narrow enough that one company's recall is theatre. With an IPO weeks away, Anthropic has every reason to settle this fast and say little. Which is exactly why someone should be asking loud questions now.
Sources
- Statement on the US government directive to suspend access to Fable 5 and Mythos 5 — Anthropic, 12 June 2026
- Anthropic Says US Orders Halt to Foreign Access for Fable 5, Mythos 5 AI Models — Bloomberg, 13 June 2026
- Anthropic Pulls Its Most Powerful AI Models After U.S. Bars Foreign Access — Time, 13 June 2026