Anthropic cuts off Fable 5 and Mythos 5 access following government order
Anthropic says it blocked all access to Fable 5 and Mythos 5 after a government order citing national security concerns. The company says it was not given specific details about the risk.
Intelligence analysis by GPT-5.4 Mini
A government export control directive forced Anthropic to cut off access to Fable 5 and Mythos 5 for all foreign nationals, including its own employees. Anthropic says it complied, but says the government did not spell out the concern and that the cited vulnerabilities were minor or already seen in other models.
Anthropic had to lock two of its AI models after the government said they could be risky for national security. It’s like a school taking away a sharp tool from everyone after being told it might be dangerous, even if the school is not told the exact reason.
Analysis
What happened
According to The Verge, Anthropic says it received a Friday evening government order requiring it to block access to Fable 5 and Mythos 5 for all foreign nationals, both inside and outside the United States. The order also applied to Anthropic employees, so the company says it had no choice but to cut off access to the models for all customers.
What Anthropic says
Anthropic says it complied with the directive, but says the government did not give it specific details about the national security issue. In its statement, the company says any evidence of a possible jailbreak was described to it verbally. It also argues that the vulnerabilities mentioned were minor and were not unique to these models, pointing to similar issues in other systems, including GPT 5.5.
The company’s defense
Anthropic says it had already taken steps to secure the models. Those steps included working with the US and UK governments and changing its data retention policy so it could better track attempts to use the models for harmful purposes. The company also says it was never given disclosure of a concerning non-universal jailbreak that actually led to harm.
Why this matters
The story shows how quickly access to advanced AI can change when governments invoke national security. It also highlights a key tension in AI policy: companies may be asked to respond to serious concerns without being told enough to fully understand or challenge them.
Key points
- A government export control directive forced Anthropic to block access to Fable 5 and Mythos 5 for all foreign nationals.
- Anthropic says the order applied even to its own employees, so it cut off access for all customers.
- The company says the government did not provide specific details about the national security concern.
- Anthropic says the reported vulnerabilities were minor and were also seen in other models, including GPT 5.5.
- The company says it had already worked with US and UK governments and changed data retention to track malicious use.
If this leads to clearer safety rules, companies and governments may work together earlier to prevent misuse of powerful AI models. Anthropic also says it had already been coordinating with the US and UK governments, which could support better safeguards going forward.
A broad access cutoff can disrupt customers and create uncertainty around who gets to use advanced AI systems. The lack of specific details from the government also leaves Anthropic unable to fully test or respond to the concern, which could keep disputes like this unresolved.
