The U.S. government just reversed course on one of the most aggressive AI export moves it has made to date. Anthropic says it will begin restoring public access to its Mythos and Fable models on Wednesday, July 1, after the Commerce Department dropped a licensing requirement that had effectively pulled the two models offline. That’s according to TechCrunch AI, which reports the reversal came after weeks of talks between the AI lab and Washington.
Here’s what actually happened, and why it matters.
What changed
On June 12, the government added Mythos and Fable to its list of export-restricted technologies. In plain terms, that meant Anthropic couldn’t make them available to foreign nationals without special approval. Policing that at scale turned out to be impractical, so the company shut off public access entirely rather than try.
Now the license requirement is gone. In exchange, Secretary of Commerce Howard Lutnick said Anthropic agreed to:
- Proactively detect and address security risks tied to the models.
- Work with the U.S. government on protocols, standards, and release procedures for Mythos, Fable, and future models.
- Report any malicious activity to the government.
Why the deal raised eyebrows
Anthropic had already pledged to do most of this voluntarily, months before the export rule existed. That timing is the whole story. Cybersecurity experts were skeptical of the restrictions from the start, and per TechCrunch AI, many read the ban less as a genuine security fix and more as leverage, a way for the Trump administration to pressure Anthropic after its executives publicly criticized how the government and the president’s political opponents might use the technology.
What stands out here is that the concessions Anthropic “agreed” to were largely commitments it made on its own before the fight even began.
The context that forced the reversal
Mythos and Fable are widely considered the most advanced AI models released so far. Mythos went to a select group of organizations back in April, in part to ease worries about its ability to find and exploit software vulnerabilities. Fable, a version with extra security guardrails, reached the public in June.
Then the competitive picture shifted. Asian AI companies, including the ones behind Fugu and Tulongfeng, started shipping models that approach Mythos-level capability. That put Washington in a bind. Restricting your most capable domestic lab while overseas rivals catch up is a hard position to defend. The pressure to let American AI compete globally is a big reason the restrictions eased.
Last week, Lutnick cleared Mythos for release to select customers approved by the White House. OpenAI’s latest models went out the same way, to a group of organizations vetted by the Trump team rather than the general public.
Why this matters for the industry
The bigger takeaway is uncertainty. The Trump administration’s approach to AI policy has been erratic, and companies across the sector now have little clarity on what rules will govern future model releases. A June executive order signaling a desire to review models before they ship drew criticism from analysts like Dean W. Ball, who recently took a policy role at OpenAI.
For practitioners, that’s the part worth watching. When access to frontier models can be switched off by a licensing decision and switched back on by a negotiation, release timelines stop being purely technical. They become political.
A few things to keep on your radar:
- Access to top-tier models may increasingly run through White House approval lists, not open release.
- “Voluntary” safety commitments can quickly become bargaining chips in export policy.
- Global competition, not just safety, is now shaping which models Americans can use and when.
Anthropic gets its models back online this week. The precedent it sets, that a frontier lab’s public access can hinge on its relationship with the administration, is going to outlast the news cycle. For the full breakdown, see the original report at TechCrunch AI.