Anthropic suspends new AI models after US security concerns

AI company Anthropic has suspended access to its newly launched models, Claude Fable 5 and Mythos 5, after US authorities raised national security concerns just days after their public release.

In a statement, Anthropic said it had been ordered to restrict foreign nationals from using Claude Fable 5, which the company previously described as its most powerful AI system to date.

“The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” the company said.

Claude Fable 5 is an advanced version of Anthropic’s Claude AI platform, competing with OpenAI’s ChatGPT and Google’s Gemini.

According to Anthropic, US officials did not identify any specific security threat but indicated concerns over a potential method of “jailbreaking” the model — a technique used to bypass built-in safeguards and uncover vulnerabilities.

The company said it reviewed demonstrations of the technique and found only a small number of previously known, minor vulnerabilities, adding that similar issues could also be identified by other publicly available AI models.

Ahead of the launch, Anthropic highlighted extensive security safeguards and initially provided limited access to selected organisations for testing, citing concerns that the model’s advanced capabilities could potentially be misused for cyberattacks or system exploitation.

The suspension comes amid increasing scrutiny of Anthropic from the Trump administration. The company is already engaged in a legal dispute with the US government over efforts to restrict federal agencies from using its AI products.

President Donald Trump has publicly criticised the company, while former Defence Secretary Pete Hegseth previously labelled Anthropic a “supply chain risk” — a designation typically reserved for entities considered unsuitable for government use due to security concerns.

Anthropic has challenged that designation in court. A federal judge has temporarily blocked its enforcement, allowing government agencies and military contractors to continue using the company’s AI tools while the case proceeds.