US imposes weapons-grade export controls on Anthropic's Fable 5 model
The US government applied export restrictions typically reserved for military hardware to Anthropic's Fable 5 model, forcing a global shutdown while the company implements geographic filtering for non-US users.
Anthropic pulled its Fable 5 model offline this week after the US government imposed export controls typically reserved for military hardware and advanced semiconductors. The restrictions require the company to block access for all foreign nationals, including Anthropic's own non-US employees. Because Anthropic cannot implement geographic filtering quickly enough, the model is now unavailable to everyone.
The government claims it discovered a jailbreak that unlocks Fable 5's capabilities for cyberattacks. Anthropic disputes the severity, stating the technique is a standard "fix this bug in the code" prompt—the same method that works in ChatGPT 5.5. The company promised additional details within 24 hours.
Dual-use policy and precedent
This marks the first time the US has applied weapons-grade export controls to a commercial AI model at scale. The government has not publicly disclosed the specific jailbreak method or provided evidence that Fable 5 poses a meaningfully higher cybersecurity risk than other frontier models already in circulation.
The timing raises questions. Anthropic is widely reported to be preparing for an IPO, and the shutdown arrives amid ongoing competition with OpenAI, which recently launched ChatGPT 5.5. Whether the restrictions stem from a genuine security concern, represent regulatory overreach, or reflect inconsistent application of existing policy remains unclear. Anthropic's promised follow-up may reveal whether the government identified a novel capability in Fable 5 or simply applied dual-use export rules to AI for the first time.
