Anthropic announced Tuesday night that the U.S. government has lifted export restrictions on Anthropic’s cutting-edge AI models.
“We have received notification that the Department of Commerce has lifted export restrictions on Claude Fabre 5 and Mythos 5,” Anthropic said in a statement. “We will begin restoring access tomorrow and will share updates soon.”
The statement came shortly after Commerce Secretary Howard Lutnick posted on social media about lifting export restrictions on Fable, one of the human models in question.
“Over the past two weeks, we have been working closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the U.S. government and strengthen U.S. leadership in AI,” Lutnick said in a post on X.
Fable is a version of Anthropic’s most advanced AI model known as Mythos, but with additional guardrails to make it suitable for public use.
The export ban and subsequent negotiations between Anthropic and the government highlight the continued regulatory uncertainty surrounding AI, even though the technology has advanced far beyond current regulations.
The government announced earlier this month that a trusted partner, which CNN has learned is Amazon, had discovered a way to jailbreak, or circumvent, Fable’s guardrails. (Anthropic said at the time that the jailbreak was “simple” and that “other publicly available models” had similar workarounds.)
The Commerce Department then issued an export ban requiring the company to suspend all use by foreign nationals, including Anthropic’s own employees, to address the risks. As a result, Anthropic has disabled customer access to both Mythos and Fable.
After Anthropic and the government worked together to address cyber risks, the Department of Commerce last week allowed Anthropic to release Mythos to select government-approved entities. Neither Anthropic nor the Department of Commerce would elaborate on what changes Anthropic made to Fable to lift the export ban.
Experts say Mythos has the potential to exploit cybersecurity vulnerabilities at an unprecedented pace. Anthropic limited the release to a group of key partners “to protect the world’s most important software.”
The White House also asked OpenAI to limit the release of the upcoming GPT 5.6 model to a small number of government-approved partners due to its advanced capabilities. “We do not believe this type of government access process should become the long-term default,” OpenAI said at the time.
