White House eases export restrictions on Anthropic's advanced AI models
The administration has lifted limits on Anthropic's Claude Mythos 5 and Claude Fable 5, allowing broader distribution after earlier orders required the models to be taken offline.
Anthropic executives met with White House officials on Monday and reached an agreement that relaxes the export controls imposed on the company's most advanced AI systems, Claude Mythos 5 and Claude Fable 5. The earlier order, issued weeks earlier, had forced Anthropic to suspend access for foreign nationals after officials said they had identified a method for bypassing the models' safeguards.
Following the talks, the administration announced that the models can now be provided to trusted organizations and, in a limited form, to the public, with Anthropic stating the public version is designed to prevent use for cyber-attacks.
The policy shift comes after the Trump administration halted a broader AI regulatory executive order, leaving agencies to determine how to manage emerging risks from powerful language models.
"The government believes it has become aware of a method of bypassing, or ‘jailbreaking’ Fable 5," Anthropic said in a blog post, prompting the initial shutdown.
The White House has also asked other AI developers, including OpenAI, to delay the rollout of newer models while it finalizes guidance on high-risk AI technologies.