Anthropic disables latest AI models Fable 5, Mythos 5 after US govt order on foreign access

Artificial intelligence startup Anthropic said on Saturday it had been ordered by the US government to suspend access to its newly launched Fable 5 and Mythos 5 models worldwide, citing unspecified national security concerns.

The company said it received the export control directive at 5:21 pm ET on Thursday and would immediately disable access to the two models for all customers.

Anthropic said the directive barred access to Fable 5 and Mythos 5 by "any foreign national, whether inside or outside the United States," including some of its own employees, effectively treating the models as export-controlled technology.

"The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance," Anthropic said in a statement.

The move comes just days after Anthropic unveiled Fable 5, which it described as its most powerful generally available AI model to date, and Mythos 5, a more capable and less restricted version available to a select group of cybersecurity and critical infrastructure professionals.

Jailbreak vulnerability

The AI startup said its understanding was that the government believed it had become aware of a way of "jailbreaking" Fable 5, allowing users to bypass some of the model's built-in restrictions.

However, the company disputed the government's assessment, saying it had reviewed a demonstration of the technique and found it identified only "a small number of previously known, minor vulnerabilities."

"We have not even received a disclosure of a concerning non-universal potential jailbreak that led to a harmful result," Anthropic said.

The company asserted that the capabilities demonstrated through the alleged jailbreak were "widely available from other models (including OpenAI's GPT-5.5)" and is used routinely by cybersecurity defenders.

Anthropic defends safety testing

The startup said it had spent thousands of hours conducting safety testing with US and British government agencies, private organisations and internal teams before launching the model.

While saying it would comply with the government's legal directive, Anthropic argued that recalling a commercial AI system over a narrow potential jailbreak sets a dangerous precedent.

"If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers," the company said.

The company said it was working with authorities and hoped to restore access to the models as soon as possible.

Fable 5 was introduced on Tuesday with a new safeguard system aimed at limiting misuse in cybersecurity and biological research. The model was designed to automatically route certain sensitive requests to an older and more restricted system, Claude Opus 4.8.

Anthropic had said the model outperformed its predecessors across software development, scientific research, data analysis and other complex tasks, while Mythos 5 offered even more advanced cybersecurity and scientific research capabilities under tighter access controls.

The company acknowledged that the systems' increased capabilities also raised new risks, prompting it to deploy specialised AI-powered monitoring tools and stricter security controls.

"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 said.