US government forces Anthropic to shut down its strongest AI models over national security concerns. The United States government has forced...
 |
| US government forces Anthropic to shut down its strongest AI models over national security concerns. |
Because Anthropic cannot reliably distinguish between foreign nationals and domestic users in real time, the practical effect of this order has been a complete global shutdown of both models for every customer worldwide. Current Fable 5 and Mythos 5 sessions end with error messages, and new queries are automatically routed to older, less capable models such as Claude Opus 4.8. Access to all other Anthropic models remains unaffected, but the shutdown represents an unprecedented intervention by Washington into the deployment of frontier AI systems.
The government's directive, which the company says was issued without specific details of its national security concern, stems from a belief that there exists a method of bypassing, or "jailbreaking," the safeguards built into Fable 5. According to Anthropic, the government expressed concern that such a jailbreak could allow the model to be used for identifying software vulnerabilities. However, the company has pushed back forcefully against this rationale, arguing that the evidence provided is insufficient to justify recalling a commercial model deployed to hundreds of millions of people.
Anthropic disclosed that the government has only provided "verbal evidence of a potential narrow, non-universal jailbreak," which essentially consists of asking the model to read a specific codebase and fix software flaws. The company reviewed the reported technique and found that the vulnerabilities it surfaced were minor and already publicly known. Furthermore, Anthropic claims that other publicly available models, including OpenAI's GPT-5.5, can discover the same flaws without requiring any jailbreak at all. No tester has yet been able to find a universal jailbreak that can broadly bypass the model's safeguards across a wide range of capabilities.
Before launching Fable 5, Anthropic had worked extensively with the US government, the UK AI Safety Institute, and multiple private third party organizations to red team the model's safeguards for thousands of hours. These tests showed that Fable's safeguards were substantially more effective than those of any previously deployed model. The company adopted a defense in depth strategy, acknowledging that perfect jailbreak resistance is not currently possible for any model provider, and implemented a controversial 30 day mandatory data retention policy for all Fable and Mythos traffic to detect patterns of misuse.
In a statement, Anthropic said, "We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people." The company warned that if this standard were applied across the industry, it would essentially halt all new model deployments for all frontier model providers. Anthropic had called for greater US oversight of AI as recently as Wednesday, including the ability to block models with unacceptable risks, but it argued that this action does not follow principles of fair, transparent, and fact based regulation.
This directive lands against a backdrop of escalating tension between Anthropic and the US government. Earlier this year, the Pentagon designated Anthropic as a national security supply chain threat after the company refused to allow the US military to use its AI models for domestic surveillance and fully autonomous weapons systems. That blacklist is set to take effect later in the year. The contradiction highlights the government's own uncertainty about how to handle a company whose AI capabilities it simultaneously fears and depends on, as the NSA continues using Claude because no alternative exists.
The action also marks a major escalation of US efforts to halt foreign adversaries' AI capabilities. For years, American export controls have focused on the chips and tools that power AI rather than on restricting foreign access to AI itself. This directive changes that approach by directly targeting the models. Dean Ball, a former White House official, said on X that the order suggests all non Americans would be restricted from using Anthropic's latest models, meaning users should expect to have to prove their citizenship to access certain AI systems in the future.
Anthropic says it believes this is a misunderstanding and is working to restore access as soon as possible. The company has promised to share more details within 24 hours. For enterprises that had integrated Fable 5 into their workflows, this sudden blackout triggers severe operational paralysis and serves as a stark warning about the risks of relying exclusively on any single cloud based frontier model provider. More information is available in Anthropic's official statement on the government directive and through ongoing Commerce Department updates on AI export controls.
