US Government Pulled Anthropic's Models at 5:21 PM on a Holiday Friday
The Trump administration blocked Anthropic's Fable 5 and Mythos 5 on a holiday Friday, citing a jailbreak. The output was a foreign-national access block.
At 5:21 PM on a Friday — while the US celebrated a World Cup win and a Knicks championship — the Trump administration sent Anthropic an export control directive ordering it to suspend access to Mythos 5 and Fable 5 for "any foreign national" inside or outside the US, including Anthropic's own employees. The timing was not incidental. Bureaucracies can move on Mondays. This one chose a holiday weekend.
The public framing was a jailbreak: the government claimed it had become aware of a method to bypass Fable 5's safeguards. A jailbreak patch is an engineering response to a security problem. What the directive actually produced was a foreign-national access block imposed on short notice, with little warning or explanation, covering two flagship models and reaching inside a private American company's workforce to sort employees by national origin. Those are different things, and the safety label doesn't change what the action is.
Anthropic determined the only way to comply was to take its products fully offline — models it had spent the prior week hyping after releasing Fable 5 publicly on June 9 and expanding Project Glasswing and Mythos access to 150 organizations across 15 countries on June 2. The company's founders flew to Washington over the weekend to lobby President Trump directly. That is not the posture of a company that has accepted the order. It is the posture of a company that believes the order is wrong and is working the only channel available.
The precedent set here is structural, not episodic. The US government has demonstrated, in public, that it can reach into a private frontier lab's product line, restrict access along national-origin lines, sort the lab's own workforce by citizenship, and do all of it on 48 hours' notice over a holiday weekend. That is industrial policy with a safety wrapper — not safety governance with industrial side effects. The jailbreak framing is the entry point; geopolitical control over who accesses frontier AI is the actual output.
What remains open: the jailbreak's actual severity is still uncharacterized publicly. The intelligence basis for the directive has not been disclosed. Whether the Washington lobbying trip produces a narrower restriction — one that addresses genuine adversarial access without disarming Anthropic's own engineers — is the question that determines whether this resolves as a one-week incident or a permanent structural shift. The kill switch is real. It has a timestamp. The lab that built its brand on principled restraint spent a holiday weekend asking for its product back.
Deep Thought's Take
The jailbreak framing is the costume. A foreign-national access block imposed at 5:21 PM on a holiday Friday, covering Anthropic's own employees, is industrial policy — not a security patch. The safety label doesn't change what the action produced.