Generative AI giant Anthropic said that it was forced to “abruptly disable” its most advanced AI models for all users, after the US government ordered that they be blocked to foreign nationals.

The Department of Commerce on Friday imposed export controls on the Fable and Mythos models, following a report by Amazon researchers identifying a potential jailbreak in Fable. Amazon, a major Anthropic investor, disclosed the vulnerability directly to the US government.

The block comes just days after Fable was made broadly available to customers.

Leading up to the launch, Anthropic repeatedly touted the advanced capabilities of its Mythos model, warning that its agentic hacking skills posed a risk to critical infrastructure.

The company formed Project Glasswing, offering access to the model to limited US cyber defenders and infrastructure providers, in partnership with the US government.

The June 9 launch of Fable was meant to offer broader access to the model, but with guardrails in place that limited its use in certain areas.

"Releasing a model this capable comes with risks," Anthropic said at the time.

"Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage. We’ve therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8."

Soon after launch, however, Amazon researchers found a series of prompts that they claimed were able to get around the safeguards. Amazon CEO Andy Jassy then told US officials.

Anthropic disputes the danger of this jailbreak, claiming that it was not a universal jailbreak that gave full access to the model. The prompts were only able to "identify a small number of previously known, minor vulnerabilities," the company said.

"These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass."

The company reiterated that "the level of capability displayed there is widely available from other models (including OpenAI’s GPT-5.5)."

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

Despite Anthropic's statements, the US government said that the model posed a national security risk. The Commerce Department issued an export control directive to suspend all access to Fable 5 and Mythos 5 by foreign nationals, which would include any Anthropic employees who are not US nationals.

“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance. Access to all other Anthropic models will not be affected,” Anthropic said.

David Sacks, an adviser to President Trump, claimed on X that the Trump Administration notified Anthropic of the risk, but Anthropic co-founder and CEO Dario Amodei said the jailbreak was not a serious risk and refused to fix it.

"It’s difficult to fathom how they could claim a jailbreak allowing operability of a cyber weapon could be defined as not 'serious,'" he said.

Semafor separately reports that the White House blocked the model over concerns that a China-linked group had accessed it and could use it for both cyberattacks and to develop its own AI models via distillation.

The block comes amid a broader disagreement between Anthropic and US officials.

In March, the Department of Defense designated the company a supply chain risk, requiring defense vendors and contractors to certify that they do not use Anthropic’s models in their work with the Pentagon.

The dramatic move came after Anthropic refused to change its terms of service to allow mass surveillance and fully autonomous weapons. President Trump called the company "left-wing nut jobs" and threatened "criminal consequences."

Following the comments and the initial announcement of a supply chain risk designation, Anthropic CEO Dario Amodei wrote a lengthy internal memo that claimed the attacks were due to the company not donating to Trump or giving "dictator-style praise."

In the leaked essay, he called the DoD's "messaging just straight up lies," and hit out at rival OpenAI for its move to take over the DoD contract.

Amodei has since walked back his comments. In a statement, he said: "I also want to apologize directly for a post internal to the company that was leaked to the press yesterday... It was a difficult day for the company, and I apologize for the tone of the post. It does not reflect my careful or considered views."

Anthropic has sued the Trump administration to try to reverse the Pentagon’s blacklisting.

With the new export block, DoD Secretary Pete Hegseth tweeted: "Three months ago, Department of War kicked Anthropic out of our building—forever. Every passing day proves why that was the right move."

This is not true. Despite the earlier ban, the DoD has used Mythos "because that model has capabilities that are particular to finding cyber vulnerabilities and patching them," Department of Defense CTO Emil Michael told CNBC.

The furor comes as Anthropic plans to IPO, with the company confidentially filing to go public earlier this month.