The US government has lifted export license requirements for Anthropic’s Mythos and Fable AI models, allowing the company to restore public access starting Wednesday, July 1. Secretary of Commerce Howard Lutnick stated the decision follows an agreement for Anthropic to proactively detect security risks and collaborate with the government on model protocols and standards.
Why were Anthropic’s AI models previously restricted?
On June 12, the US government added Mythos and Fable to its list of export-restricted technologies. This designation prohibited the models from being made available to foreign nationals without obtaining special government approval. Anthropic reported that complying with these regulations proved impractical at scale, which forced the company to end all public access to the models.
Mythos was initially made available to a specific group of organizations in April to address concerns regarding its ability to identify and exploit software vulnerabilities. A version called Fable was subsequently released to the public in June, featuring additional security guardrails.
What did Anthropic agree to in order to restore access?
Following several weeks of discussions, Secretary of Commerce Howard Lutnick announced that Anthropic reached an agreement with the government to ensure security. According to Lutnick, Anthropic has agreed to:
- Proactively detect and address security risks associated with the models.
- Work diligently with the U.S. government on protocols, standards, and releases for Mythos, Fable, and future models.
- Inform the U.S. government of any malicious activity detected.
Cybersecurity experts had expressed skepticism regarding the original restrictions. Some analysts suggested the ban functioned as leverage rather than a security necessity, noting that Anthropic had already pledged to implement many of these safety measures voluntarily months before the export rule was enacted.
How is global competition driving US AI policy?
The decision to ease restrictions comes as American companies face increasing competition from Asian AI developers. Companies such as Fugu and Tulonfeng have begun releasing models that approach the capabilities of Anthropic’s Mythos.
To ensure American AI remains competitive in the global market, the US government moved to lift the barriers that had previously limited Anthropic’s reach. This shift highlights the tension between maintaining national security through export controls and the economic necessity of leading the global AI sector.
Comparing AI Release Strategies
Current US policy shows a divergence in how major AI labs are permitted to distribute their technology:
| Company/Model | Release Method |
|---|---|
| Anthropic (Mythos/Fable) | Restoring general public access. |
| OpenAI (Latest Models) | Released to organizations approved by the Trump administration. |
| Mythos (Select Customers) | Cleared for customers approved by the White House. |
What happens next for AI regulation in the US?
The Trump administration’s approach to AI policymaking has created a landscape of uncertainty for industry leaders. An executive order issued in June, which signaled a desire to review models prior to their release, drew criticism from industry analysts.
Dean W. Ball, who holds a policy position at OpenAI, has been among the voices criticizing the direction of these regulatory efforts. The industry continues to monitor how future model releases will be governed as the government balances security protocols with the need for rapid technological deployment.
Frequently Asked Questions
When will Anthropic’s Mythos and Fable be available again?
Anthropic stated it will begin restoring access to these models on Wednesday, July 1.
Why were the models restricted in June?
The US government placed them on an export-restricted list to manage the security risks associated with their advanced capabilities, particularly regarding software vulnerability exploitation.
Is all AI access being restricted to approved groups?
Not necessarily. While Anthropic is restoring public access, other companies like OpenAI have released their latest models specifically to groups approved by the administration.
Stay informed on the latest in AI regulation and technology. Comment below with your thoughts on these policy shifts or subscribe to our newsletter for daily updates.