Anthropic withdraws access to powerful AI model after US government order | Science, Climate & Tech News

Anthropic has withdrawn its powerful Fable 5 model after the US government ordered the company to revoke access to all foreign nationals.

The Trump administration cited “national security authorities” and told the company to “suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees”, according to Anthropic.

As a result, all customer access to the models has been “abruptly” disabled to ensure compliance, it said. Access to Anthropic’s other models is still available.

Share

What risks do AI models such as Mythos pose?

Although the government did not give Anthropic “specific details” about its concerns, the company said it believes the order comes from a “misunderstanding” over a “narrow” jailbreak.

Jailbreaking is when users manage to bypass software restrictions protecting a cyber network.

Before it was released, Anthropic itself described Mythos, Fable 5’s base model, as “too powerful”, saying it had an exceptional ability to find security vulnerabilities in software.

As a result, 50 companies were given access to the AI in order to test their systems for vulnerabilities before the public was allowed access.

Share

Will AI kill coding jobs?

Advertisements

Three days ago, Anthropic released Fable 5, which is a heavily-guardrailed version of Mythos, to the public.

Immediately, it showed itself to be the most capable AI model available to the public, according to benchmark tests from Vals AI, an AI performance tracker.

Read more on Anthropic:
Anthropic warns of ‘risks of humans losing control over AI’
The AI delusion problem is bigger than we thought

Now, Anthropic says it believes the safeguards it has in place on Fable “reduce the likelihood that Fable is misused for tasks related to cybersecurity”.

It added: “In fact, our safeguards are so strong that many users have complained that they are overly broad.”

Share

Pope: AI must serve humanity not concentrate power

It said it has reviewed a report that it believes the government has based its directive on and found the jailbreak could be deployed by many other models, including OpenAI’s GPT-5.5.

“We believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts,” said the company in a statement.

“This action does not adhere to those principles.”

The UK’s AI minister, Kanishka Narayan, said in a post on X the main lesson to take from the withdrawal is that “access to AI capabilities is crucial” as we “debate the future of national security and technology sovereignty”.