Anthropic announced Tuesday that the U.S. Department of Commerce has lifted export restrictions on its Fable and Mythos AI models, giving the tech giant a major boost to the competitive landscape.
The announcement comes after the Trump administration ordered the U.S.-based artificial intelligence company to suspend public access to Anthropic’s Power Mythos class AI models, citing national security concerns.
The Trump administration has adopted an aggressive surveillance strategy for Frontier AI models, citing concerns that the advanced systems could be exploited by military intelligence agencies in countries such as China and Russia.
In a recent development, Anthropic has reached an agreement with the Trump administration to restore access to Fable, a public version of the powerful Mythos.
“We have received notification that the Department of Commerce has lifted export restrictions on Claude Fable 5 and Mythos 5. We will begin restoring access tomorrow,” Anthropic said in an X post.
U.S. Secretary of Commerce Howard Lutnick sent a letter to Anthropic noting the lifting of the restrictions, but adding that Anthropic “has agreed to proactively detect and address security risks associated with its models.”
Lutnick said the department “reserves the right to reevaluate the decisions made in this letter and the need to reimpose licensing requirements if circumstances change or if Anthropic fails to honor its commitments.”
Additionally, as part of the agreement, Anthropic will work with the U.S. government to set protection protocols for Fable, Mythos, and future models. The company will report any violations or malicious acts to management.
As reported by the Wall Street Journal, Anthropic has implemented new safeguards to block workarounds identified by Amazon researchers that previously allowed users to circumvent safety guardrails and conduct cyberattacks. Anthropic claims this new measure will stop 99% of such evasion attempts.
Additionally, Anthropic is working with Google, Microsoft, and Amazon to develop a “consensus framework” aimed at defining and measuring the critical nature of “jailbreaks.” It also standardizes how developers should respond.
The lifting of this restriction comes after the US last week allowed Anthropic to release Claude Mythos 5 AI to over 100 trusted US organizations.
Security experts remain concerned. Isaac Harris, executive director of the Frontier Security Institute, a nonprofit focused on AI and national security, questioned the United States’ ability to handle technology with unprecedented capabilities coming from China.
