US Removes Export Restrictions on Anthropic's Fable 5 and Mythos 5 Models
The Commerce Department lifted export curbs on Anthropic's latest AI models, allowing global access to Fable 5 and restoring US access to Mythos 5 after a safety review and new safeguards were implemented.
The U.S. Commerce Department has ended export restrictions on Anthropic's newest Claude models, Fable 5 and Mythos 5, permitting worldwide deployment of Fable 5 and reinstating access to Mythos 5 for U.S. organizations as of June 26. Anthropic announced the changes in a blog post and said it is expanding the Glasswing program, which grants trusted cybersecurity researchers access to Mythos for defensive work.
In a letter viewed by Reuters and The New York Times, Commerce Secretary Howard Lutnick stated that Anthropic will no longer require a license for exporting or transferring the models within the United States, noting that the company had coordinated with the government to mitigate identified risks. Lutnick added that the United States "reserves the right to re-evaluate the decisions" and could reimpose curbs if necessary.
The restrictions, first imposed on June 12, were prompted by concerns that the models could be exploited by foreign actors to target critical infrastructure. Anthropic responded by shutting down all external access and developing additional safeguards, including a red-team program and a 24/7 monitoring team for jailbreak threats. The company reported that a vulnerability-exploitation method identified by Amazon researchers has been blocked in more than 99 % of cases, though it acknowledged that the enhanced safety classifier may occasionally block benign coding requests.
Anthropic is also collaborating with industry partners such as Amazon, Microsoft, Google and other Glasswing members to create a framework for assessing AI jailbreak severity. The firm launched a HackerOne program to encourage security researchers to submit potential jailbreaks for Fable 5.
The renewed partnership follows a prior legal dispute in which Anthropic challenged a national-security designation that it said was retaliation for refusing government requests related to autonomous weapons and mass surveillance. The company now plans to provide early access to frontier models for government testing and to share information on emerging threats, urging Congress to enact legislation that aligns regulatory standards for AI developers.