Your AI Stack Has a Kill Switch

On Friday at 5:21 PM Eastern, one of the most capable AI models a company could access stopped being available. Not because of an outage or a flaw Anthropic found. Because the US Commerce Department sent a letter. The order, citing national security, told Anthropic to suspend access to its two newest models for any foreign national anywhere, including its own employees. The only way to comply was to shut both models off for everyone, everywhere. A model that had been live that morning returned a 404 by dinner, with no restoration date. The uncomfortable translation for anyone who spent the last year moving real work onto a frontier model: the thing you built on can be switched off by a party that is not you, not your vendor, and not answerable to your roadmap. Treating frontier model access as a utility stopped being a safe assumption on Friday at 5:21 PM. Single-model dependency is now a continuity risk with a name, and the operators who come out fine are the ones who already know what breaks when the demo disappears.

The full issue, slide by slide