The White House has no plan to expand export controls imposed on Anthropic’s latest model to other AI companies, The Information reported on June 13 local time.

A U.S. government official said the move stemmed from Anthropic’s refusal to fix vulnerabilities in its Fable5 and Mythos5 models. David Sacks (데이비드 삭스), the former White House AI and cryptocurrency czar, also said on social media that Dario Amodei (다리오 아모데이), Anthropic’s CEO, rejected requests to fix vulnerabilities or halt the model.

Anthropic said other companies also have models with similar security risks, and pointed to OpenAI’s GPT-5.5, released in April, as being at a similar level to part of the Mythos cybersecurity benchmark. In a statement, Anthropic said a "narrow range of potential jailbreaks" should not justify an order to pull commercial models. It said disclosed potential jailbreaks were "completely harmless or minor."

The government was concerned that Fable, which was made available to a broad range of developers, could find vulnerabilities in sensitive systems and that advanced AI capabilities could spread to hostile forces. Under the export-control directive, Anthropic blocked access to its latest model for all customers as well as foreign nationals.