The United States has lifted export restrictions on Anthropic’s most advanced artificial intelligence models, allowing the company to restore access to Claude Fable 5 and Mythos 5 after weeks of government-imposed limits tied to national security concerns.
Anthropic said it will begin restoring access to the models on Wednesday after the US Department of Commerce informed the company that export controls had been removed.
The restrictions had been introduced shortly after the release of the two models in June, with the government citing security concerns linked to their advanced capabilities.
According to a letter from Commerce Secretary Howard Lutnick, Anthropic no longer requires an export licence after agreeing to proactively detect and address security risks associated with the models.
The company also agreed to work with the US government on standards for future AI models and to report malicious activity involving the technology.
The Commerce Department said it reserves the right to reconsider its decision if necessary.
Anthropic previously said authorities had not identified a specific issue beyond concerns about a possible method of bypassing the safety protections in Fable 5.
The company had maintained that the reported vulnerability did not justify withdrawing a commercial model already deployed to users.
Last week, Anthropic said it had already received approval to provide the models to US organisations responsible for operating and defending critical infrastructure while discussions continued over broader public access.
The decision follows a period of tension between Anthropic and the Trump administration over AI oversight and national security.
The company had also challenged the US Department of Defense earlier this year after being designated a supply chain risk.
The latest move comes as the administration has increased scrutiny of advanced AI systems despite initially favouring a lighter regulatory approach.
OpenAI also recently announced that its GPT-5.6 series would initially be released only to a limited group of trusted partners following government pressure to stagger deployment.
Analysts said the reversal had been widely anticipated and could shape future interactions between frontier AI developers and US regulators, particularly regarding whether government approval could become a standard part of releasing advanced AI models.