Claude Mythos 5: USA partially lifts ban Apfelpatient

Nearly two weeks after the surprise export ban, the US government has partially relented: Anthropic is allowed to resume shipping its most powerful AI model, Claude Mythos 5 – but only to a select group of over 100 institutions. The situation for its sister model, Claude Fable 5, remains unclear for now.

In mid-June, an export control directive from Washington forced Anthropic to cut off access to the Claude Fable 5 and Mythos 5 for all customers. Since then, the company and the government have been negotiating a solution. These talks have now yielded an initial result: According to several reports, the US government has partially lifted the restrictions and is re-enabling the more powerful of the two models for a defined group of recipients.

What the government now allows

Specifically, Anthropic is permitted to deliver Claude Mythos 5 to more than 100 US institutions, including large corporations and government agencies – reportedly numerous Fortune 500 companies. The decision is documented in a letter from Commerce Secretary Howard Lutnick to Tom Brown, Anthropic's Chief Compute Officer. In the letter, Lutnick explains that appropriate safeguards have been put in place to allow access to the model for selected trusted partners and points to significant progress in the intensive, daily discussions since the embargo took effect.

For the recipients listed by name in an appendix to the letter, a separate export license will no longer be required. In return, Anthropic has committed to working with the US government on protocols and standards for future model releases.

Claude Fable 5 will be left out for now

The situation is different with the second model. Claude Fable 5 – the variant that was the first publicly available model in the Mythos class – is not mentioned in the letter. It remains unclear when, or even if, customers will be able to access it again. While sources close to the negotiations indicate that a release is also imminent, there is no timeline.

How the ban came about

The conflict was triggered by a directive on June 12 that forced Anthropic to disable both models worldwide for all users. The exact reasons remain unclear. According to most reports, the government became aware of the issue after Amazon CEO Andy Jassy alerted US Treasury Secretary Scott Bessent to a suspected jailbreak that allegedly made the models vulnerable to abuse. There were also indications that the White House was concerned about China's potential access to Claude Mythos. In the following two weeks, both sides worked to resolve the dispute.

A new regime for top-of-the-line models

The partial release marks a significant easing of tensions between the government and one of the leading AI labs. At the same time, it indicates how the handling of particularly powerful AI models is changing: The state is increasingly demanding a say in who has access to such systems. The fact that the news arrived just a few hours after the strictly limited release of OpenAI's new model GPT-5.6, which is also initially only available to a small, government-approved group, underscores this trend. For the broader customer base of both providers, the question remains, for now, when their most powerful models will be fully available again. (Image: Shutterstock / Primakov)