Anthropic said it has shut down access to its Claude Fable 5 and Claude Mythos 5 models worldwide after receiving an order from the U.S. government citing national security concerns.
The company said the directive was received on Friday at 5:21 p.m. ET and that it complied immediately.
According to Anthropic, access to its other AI models remains unaffected.
Global shutdown
Although the government action was presented as an export-control measure aimed at limiting foreign access, Anthropic said the order required it to disable both models for all users globally.
The company said it believes the concerns stem from an alleged jailbreak involving Fable 5.
Mythos 5 was tightly controlled
Anthropic previewed Claude Mythos 5 in April and had restricted access because of its advanced cybersecurity capabilities.
According to the company, the model successfully identified vulnerabilities in every major operating system and web browser it tested.
Rather than releasing it publicly, Anthropic created Project Glasswing, providing access to around 50 vetted organizations, including:
- Amazon.
- Apple.
- Google.
- Microsoft.
- CrowdStrike.
The company said those organizations used the model for defensive cybersecurity purposes.
Fable 5 launched three days earlier
Claude Fable 5, introduced earlier this week, was designed as a more broadly accessible version of Mythos 5.
Anthropic said the model included safeguards intended to prevent responses involving high-risk areas such as:
- Cybersecurity.
- Biology.
According to benchmark tests from Vals AI, Fable 5 was the highest-performing publicly available AI model.
Anthropic disputes the government’s concerns
In a blog post, Anthropic said authorities have so far provided only verbal evidence of what it described as a “potential narrow, non-universal jailbreak.”
The company said the issue involved prompting the model to analyze a specific codebase and identify software vulnerabilities.
Anthropic argued that similar capabilities are already available in competing systems, including OpenAI’s GPT-5.5, and are commonly used by cybersecurity professionals.
The company also said its most important safeguards rely on separate classifier systems that continue to block dangerous outputs even if users bypass refusal mechanisms.
“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,” Anthropic said.
“If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”
Safety reputation under scrutiny
Anthropic has built much of its reputation around AI safety and has often positioned itself as a more cautious alternative to competitors.
The company is widely expected to pursue an initial public offering this year.
Altman previously criticized Anthropic’s approach
OpenAI chief executive Sam Altman criticized Anthropic’s handling of Mythos 5 in April during an interview with podcaster Ashlee Vance.
Altman described the strategy as “fear-based marketing.”
“It is clearly incredible marketing to say, ‘We have built a bomb. We were about to drop it on your head. We will sell you a bomb shelter for $100 million,'” Altman said.
According to the report, observers have noted that Anthropic’s repeated warnings about the capabilities of Mythos 5 may have contributed to the government scrutiny now facing the company.
TL;DR
Anthropic said it disabled its Claude Fable 5 and Mythos 5 models worldwide after receiving a U.S. government order tied to national security concerns. The company disputes the decision and says the standards being applied could affect AI development across the industry.
AI summary
- Anthropic shut down Fable 5 and Mythos 5 worldwide.
- The action followed a U.S. government directive.
- Mythos 5 had been restricted to vetted organizations.
- Anthropic disputes claims related to an alleged jailbreak.
- Sam Altman previously criticized Anthropic’s messaging around the model.
