Trump lifts export restrictions on Anthropic Mythos and Fable AI models

Trump lifts export restrictions on Anthropic Mythos and Fable AI models

The Trump administration removed export restrictions on Anthropic’s Claude Fable 5 and Mythos 5 artificial intelligence models on Tuesday evening, ending a weeks-long dispute over the security of the software. The move allows Anthropic to restore access to its most advanced AI tools for users in the U.S. And abroad.

Anthropic announced it would begin restoring access to the models on Wednesday, July 1. Access to Fable 5 will be restored globally across several platforms, with redeployment on Google Cloud and Amazon's AWS to follow as quickly as possible.

The conflict began on June 12, when the U.S. Government added the products to a list of export-restricted technologies. This required Anthropic to obtain licenses before providing access to foreign nationals, a requirement the company said made public access impractical, forcing a total shutdown of the models.

The terms of the agreement

Commerce Secretary Howard Lutnick informed Anthropic in a letter that a license is no longer required for the export, reexport, or in-country transfer of the models. According to Lutnick, the decision follows a national security review by the Bureau of Industry and Security.

To secure the lift of these restrictions, Anthropic agreed to several conditions:

• Proactively detect and address security risks associated with the models.
• Work with the U.S. Government on protocols, standards, and releases for Mythos, Fable, and future models.
• Inform the government of any detected malicious activity.

Anthropic is also redeploying Fable 5 with new classifiers designed to block cybersecurity tasks. The company noted that some routine debugging and coding tasks will temporarily fall back to Opus 4.8 while these classifiers are refined to reduce false positives.

National security and global competition

The administration's restrictions were driven by concerns that the models could be used to identify and exploit software vulnerabilities. Anthropic initially disagreed, arguing that the government's concerns centered on a narrow potential jailbreak that was relatively simple and replicable in other models.

However, pressure mounted on the White House as Asian AI companies, including Tulonfeng and Fugu, began releasing models with capabilities approaching those of Mythos. Glenn Gerstell, a senior adviser at the Center for Strategic & International Studies and former NSA general counsel, stated that the U.S. Cannot maintain a lead over China by slapping export controls on the latest innovation.

Mythos 5, which possesses fewer guardrails than Fable 5, was originally released in April to a select group. While the general export ban was lifted Tuesday, Anthropic stated it will continue coordinating with the government to expand Mythos access to select partners.

A history of friction

The export ban is the latest clash in a volatile relationship between the AI lab and the Trump administration. Earlier this year, Anthropic sued the Department of Defense after Defense Secretary Pete Hegseth labeled the company a supply chain risk and ordered federal agencies to stop using its technology. This followed a dispute over guardrails to prevent the military from using Claude for mass surveillance or autonomous weapons. A federal judge blocked those restrictions, describing them as Orwellian, though the government is appealing.

Some cybersecurity experts suggested the June export ban was less about security and more about leverage against Anthropic executives who had criticized the administration. Reports indicate the atmosphere shifted after Anthropic cofounder Tom Brown replaced CEO Dario Amodei in government meetings, as officials reportedly preferred Brown on a personal level.

Industry-wide implications

The case has created anxiety across the AI sector regarding the predictability of U.S. Policymaking. Last week, OpenAI similarly limited the release of its GPT-5.6 model to a small group of approved organizations following White House pressure, despite objections from CEO Sam Altman.

White House chief of staff Susie Wiles defended the administration's approach in a Tuesday post on X, citing a June 2 executive order emphasizing voluntary federal reviews. She stated the priority is to get the best tech deployed as quickly and safely as possible.

Industry advocates remain skeptical. Paul Lekas of the Software & Information Industry Association emphasized the need for a consistent process for model assessment. Joe Hoefer, chief AI officer of Monument Advocacy, described the resolution as more of a ceasefire than a full solution, noting that the lack of a codified process means future launches carry the risk of similar interventions.

Anthropic is now working with Microsoft, Google, and Amazon to develop a framework for assessing AI jailbreaks and determining developer responses.