In an unprecedented move, the US government has ordered artificial intelligence company Anthropic to suspend worldwide access to its flagship AI models, Fable 5 and Mythos 5, citing national security concerns over a possible method of bypassing the systems’ built-in safeguards.
The order effectively forces the company to take its most advanced AI products off the market, affecting users across the world. Anthropic said access to all its other AI models would continue uninterrupted.
In a statement, the company said it received the directive from US authorities and immediately began disabling the two models to comply with the order. However, Anthropic disputed the basis for the action, saying the government had not provided detailed evidence to justify the shutdown.
According to the company, officials are concerned about a potential “jailbreak”—a technique used to trick an AI system into ignoring some of its safety restrictions. Anthropic said it reviewed the method in question and found that it revealed only a handful of previously known and relatively minor software vulnerabilities.
The company maintained that the capability demonstrated was not unique to Fable 5 and could be replicated using other publicly available AI models. It also said there was no evidence that the alleged flaw had caused any real-world harm.
Anthropic said Fable 5 had undergone extensive safety testing involving internal teams, outside experts, the US government and Britain’s AI Safety Institute before its release. The company claimed those tests showed its safeguards were stronger than those of previous generations of AI models.
It acknowledged, however, that no AI system is completely immune from attempts to circumvent its protections. To address such risks, the company said it had adopted multiple layers of safeguards and monitoring measures to detect and contain misuse.
Anthropic said the government had so far shared only verbal descriptions of what it considered a limited and highly specific jailbreak method, involving requests to analyse computer code and identify flaws. The company said the same capabilities are routinely used by cybersecurity professionals and are already available through other AI systems.
“We are complying with the government’s legal directive and are removing access to Fable 5 and Mythos 5 for all users,” Anthropic said. “However, 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 similar standards were applied across the industry, they could effectively halt the release of future frontier AI systems.
Anthropic said it supports government oversight of unsafe AI technologies, but argued that such interventions should be based on a transparent process backed by clear technical evidence.
Calling the move a “misunderstanding”, the company apologised to customers for the disruption and said it was working to restore access to the models as quickly as possible.
