Already a subscriber? Make sure to log into your account before viewing this content. You can access your account by hitting the “login” button on the top right corner. Still unable to see the content after signing in? Make sure your card on file is up-to-date.

The US government has loosened the restrictions it slapped on Anthropic’s most advanced AI model, clearing the company to give more organizations access to Claude Mythos 5.

Some shit you should know before you dig in: If you’re unaware, the US Commerce Department abruptly ordered Anthropic on June 12 to disable both Mythos 5 and its consumer-facing counterpart, Fable 5, citing national security concerns and giving the company just 90 minutes to pull them offline. The directive required Anthropic to block all foreign nationals from using the models, including its own foreign national employees and people living and working in the US, so the company shut off access entirely. The administration’s worries reportedly traced back to a paper from Amazon researchers who got the model to disclose flaws in vulnerable software code, raising fears it could be weaponized for cyberattacks. This came shortly after the Trump administration and Anthropic clashed over a $200 million Pentagon contract, a fight over how the tech should get deployed on the battlefield, which led the Defense Department to label Anthropic a “supply chain risk” (a designation meaning the company posed a national security threat, one never before aimed at a US company). Anthropic sued over the label in March, and that litigation is still ongoing.

What’s going on now: In a letter to Anthropic co-founder Tom Brown, Commerce Secretary Howard Lutnick said the company had done enough to bring Mythos 5 back for a limited group of clients. He wrote, “I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model.”

The secretary said Anthropic had worked with the government to address the risks tied to the models and that those efforts had “yielded significant progress.” He also flagged that he can reshuffle who’s on the approved partner list “at any time” and reserved the right to revisit the license requirements if circumstances change. The approved organizations can now hand Mythos access back to their foreign national employees, and Anthropic gets the same clearance for its own.

The deal does not cover Fable 5. Lutnick’s letter said nothing about restoring access to the consumer version, and the rest of the requirements from his original June 12 directive remain in place. Some outlets reported that in the background Anthropic was still in talks with the White House over Fable 5, with both sides hoping the whole mess ends up shaping how future model releases get handled.

This all comes as the administration moves to take a more hands-on role in how AI models get released. The same day Lutnick sent his letter, rival OpenAI said it would limit the rollout of its new GPT-5.6 models to a small group of government-approved partners, calling it a “short-term step” it didn’t want to see become the long-term default.

The White House also called on Meta this week to turn its models over for a voluntary check, which leaves the company as the only major US AI developer still refusing to play along.

These calls have led critics to warn that the government’s approach is becoming increasingly restrictive, with former White House AI adviser Dean Ball writing that, in just a few weeks, federal AI policy has shifted “from implausibly libertarian to increasingly draconian and opaque.”