AI Governance
Data Centers
Drones
Climate
Privacy
2026 Maine Senate
Bob Dylan
Commerce Forces Anthropic Models Offline
Coverage from Spencer Fane, Startup Fortune, and others
The U.
S. Commerce Department required Anthropic to obtain authorization before sharing its Fable 5 and Mythos 5 models with foreign nationals, citing national-security concerns related to a reported jailbreak. Anthropic could not reliably verify users’ nationality in real time and therefore disabled the models worldwide, affecting customers and employees beyond the targeted group. The restrictions were later withdrawn after Anthropic reported stronger safeguards and agreed to expanded government testing and information sharing, but Commerce retained the ability to reinstate them. The episode demonstrates that export-control powers can directly affect frontier-model availability without new AI-specific legislation, leaving companies and customers exposed to abrupt and potentially nontransparent interventions.
The account is materially unchanged; the main refinement is that Commerce withdrew the restrictions on June 30, 2026, confirming a temporary and conditional resolution.
The biggest change is that the Commerce action was not just imposed but later rolled back after Anthropic added safeguards and accepted ongoing government collaboration. That turns the story from a one-way restriction into a conditional regulatory settlement with a new precedent for ad hoc control of deployed AI models.
- Commerce later withdrew the controls after Anthropic added safeguards.
- Anthropic accepted deeper pre-release testing and information-sharing commitments.
- The shutdown lasted 18 days globally.
- Commerce can reimpose the licensing requirement later.
- The case is framed as precedent for ad hoc AI export controls.
The main update is a reframing of the dispute: the Commerce Department’s order is now described more explicitly as a national-security restriction on frontier models, and Anthropic’s response is characterized as a broad workaround that caused a global interruption. The current version also more clearly raises the stakes by suggesting the order could slow or halt future frontier releases if applied broadly.
The U.S. Commerce Department ordered Anthropic to restrict access to its most advanced AI models for foreign nationals, prompting the company to disable them worldwide because it could not reliably enforce nationality-based limits. Anthropic says the government’s concern appears to center on a narrow jailbreak rather than a total safety failure, and it has argued similar capabilities exist in rival models. The episode marks an unusual escalation in AI governance by using export-control authority directly against model access, not just chips or compute.