By Chris Thomas, Karen Freifeld and Deepa Seetharaman
June 30 (Reuters) - Anthropic said on Tuesday that the U.S. Commerce Department lifted export controls on its Fable and Mythos AI models, less than three weeks after the company was ordered to suspend access to its most advanced AI models over national security risks.
Washington has stepped up oversight of new model releases to identify potential threats amid concerns that advanced AI models could be misused by military intelligence in China, Russia
or other countries of concern.
Anthropic disabled its Mythos 5 and Fable 5 models following a June 12 export-control order. Last week, the U.S. government allowed the company to release Mythos 5 but only to some "trusted" U.S. organizations.
"We've received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5. We'll begin restoring access tomorrow," Anthropic said in an X post.
A letter to Anthropic from U.S. Commerce Secretary Howard Lutnick that was seen by Reuters also said the export controls were withdrawn.
Anthropic had "agreed to proactively detect and address security risks associated with the models," Lutnick said. It had also agreed to work diligently with the U.S. government on protocols for Mythos, Fable and future models and to inform the U.S. government of any malicious activity, he added.
US GOVERNMENT MIGHT REEVALUATE IF CIRCUMSTANCES CHANGE
Lutnick, however, said that the department "reserves the right to reevaluate the decisions made in this letter and the necessity of reimposing a license requirement, should circumstances change or should Anthropic fail to adhere to its commitments".
Both Fable and Mythos use the same underlying AI model, but Fable is designed to be widely available for public use whereas some safeguards are lifted for Mythos.
Anthropic has implemented a new safeguard that targets and blocks a technique that it believes the government viewed as a method of bypassing or "jailbreaking" Fable 5, a company source said, declining to be identified.
The source, however, added that the vulnerabilities found by the technique which had concerned the government had been known and patched earlier.
Experts have said that Mythos models, in the wrong hands, could dramatically accelerate sophisticated cyberattacks, particularly in sectors such as banking that rely on complex, interconnected, and often decades-old technology systems.
The government's decision on Friday to allow access to Mythos to some "trusted" U.S. organisations came in tandem with an announcement by rival OpenAI that it had delayed a full public launch of GPT-5.6 at the U.S. government's request, limiting its access to a small group of vetted partners.
Those government actions had drawn much criticism that the administration's choosing of which companies were worthy of being considered "trusted" was unfair and not transparent.
Increased scrutiny of AI models this month began with U.S. President Donald Trump's signing of an executive order establishing a voluntary framework for AI developers to offer "covered frontier models" to the U.S. government for up to 30 days before releasing them to trusted partners.
Isaac Harris, executive director of the Frontier Security Institute, a nonprofit focused on AI and national security, said that Tuesday's lifting of curbs indicated that "there’s now a process for standards of the U.S. models."
But he added: "There's still a question mark as to how equivalently dangerous capabilities coming from China with less guardrails will be handled by the administration in the U.S. market."
Both OpenAI and Anthropic have confidentially filed for U.S. initial public offerings.
Anthropic's relationship with the U.S. government has been particularly rocky this year. The Pentagon designated the company a "supply-chain risk", preventing contractors from using Anthropic's AI when working for the U.S. military, after the company refused to allow its models to be used for mass domestic surveillance or fully autonomous weapons systems.
(Reporting by Chris Thomas, Karen Freifeld and Deepa Seetharaman; Additional reporting by Alexandra Alper, Costas Pitas, Ismail Shakil and Jaspreet Singh; Editing by Paul Thomasch, Rashmi Aich, Stephen Coates and Edwina Gibbs)
