Anthropic has suspended access to its Fable 5 and Mythos 5 AI models following a directive from the US government issued under national security authorities.
According to the company, the order requires it to block access to the two models for all foreign nationals, regardless of whether they are located inside or outside the United States. The restrictions also extend to foreign national employees working at Anthropic.
Access Pulled Worldwide
To comply with the directive, Anthropic said it was forced to abruptly disable Fable 5 and Mythos 5 for all customers. The company clarified that access to its other Claude models remains unaffected.
Anthropic apologised for the disruption, describing the situation as a misunderstanding and stating that it is working to restore access as quickly as possible.
Why Did the US Government Step In?
The company said it received the government’s notice on June 12 at 5:21 p.m. ET. However, the directive reportedly did not provide detailed information about the specific national security concern that prompted the action.
Based on its understanding, Anthropic believes authorities were concerned about a potential method to bypass, or “jailbreak,” Fable 5’s safeguards.
After reviewing demonstrations of the technique, Anthropic said it identified only a small number of previously known software vulnerabilities. The company argued that these weaknesses were relatively minor and could likely be discovered using other publicly available AI systems without relying on the alleged jailbreak.
What Are Fable 5 and Mythos 5?
Anthropic introduced Claude Fable 5 on June 9 as its most advanced AI model designed for broad commercial use. The company highlighted its capabilities across software engineering, scientific research, visual tasks, knowledge work, and complex long-duration assignments.
At the same time, Anthropic launched Claude Mythos 5 through Project Glasswing, providing access to a limited group of cybersecurity professionals and infrastructure providers.
Mythos 5 is built on the same underlying technology as Fable 5 but offers trusted users greater flexibility by relaxing certain safeguards in selected areas.
Anthropic Defends Its Safety Measures
The company stressed that Fable 5 includes extensive protections aimed at preventing misuse, particularly in cybersecurity-related scenarios.
Before the model’s release, Anthropic said it collaborated with the US government, the UK AI Security Institute, third-party organisations, and its own internal teams to conduct thousands of hours of safety testing.
According to the company, none of the testers identified a universal jailbreak capable of broadly bypassing the model’s safeguards.
Anthropic added that it uses a “defense-in-depth” strategy designed to make successful jailbreak attempts difficult, costly, and easier to detect through continuous monitoring.
Industry Pushback Grows
Reuters reported that senior Anthropic engineers were scheduled to meet officials from the US Department of Commerce in Washington to discuss the matter. The report also said the company had been in frequent contact with government representatives since receiving the directive.
The restrictions have sparked concern among cybersecurity experts. According to Axios, several industry leaders argued that limiting access to advanced AI systems could ultimately weaken defenders more than malicious actors.
They noted that similar capabilities already exist in competing AI models and warned that pulling Fable 5 from the market could create uncertainty without significantly reducing security risks.
Anthropic Says the Decision Sets a Dangerous Precedent
While confirming that it is complying with the legal order, Anthropic said it disagrees with the government’s approach.
The company argued that a narrow and potentially manageable vulnerability should not justify the withdrawal of a commercial AI model used by a broad customer base.
Anthropic warned that applying the same standard across the AI industry could severely slow innovation and disrupt future launches of advanced AI systems.
“The government should have the ability to stop genuinely unsafe deployments,” the company said, “but such decisions should be based on transparent processes and clear technical evidence.”
According to Anthropic, the current action falls short of that standard.
