Washington Eases Up on Anthropic, Clears Mythos 5 for Release to Select Partners

Anthropic just got some breathing room from Washington.

The U.S. government said Friday that it's letting Anthropic release its advanced Claude Mythos 5 model to a group of "trusted partners."

The decision comes just two weeks after the government forced the company to pull access to some of its models, over worries that military or intelligence groups in countries like China and Russia could get their hands on them.

Who Gets Access Now

More than 100 companies and institutions, including a long list of Fortune 500 names and several federal agencies, will now be able to use Mythos 5 again, according to people familiar with the decision. Anthropic and the White House haven't commented publicly yet.

This whole episode started after Anthropic suddenly took both Mythos 5 and its sibling model, Fable 5, offline for every single user. That happened right after the government issued its export control order.

The shutdown caught a lot of people off guard, since these were Anthropic's most advanced models at the time, and businesses around the world had built workflows around them.

What the Commerce Secretary Said

Commerce Secretary Howard Lutnick laid out the decision in a letter sent directly to Anthropic. He said that since his original order on June 12, Anthropic had worked closely with the government to deal with the risks tied to these models, and that the work had paid off in a meaningful way.

Lutnick also spelled out exactly what's changing. Companies that make the approved list, along with their foreign employees and Anthropic's own foreign staff, will no longer need an export license to access Mythos 5. But for any company that didn't make that list, the restrictions stay exactly where they were.

A Commerce Department spokesperson framed the speed of it all as a win, saying the government moved fast over the past two weeks to keep America ahead in AI while still keeping security front and center.

Worth noting: Lutnick didn't leave himself boxed in. He made clear in the letter that he can reconsider and adjust these license rules at any time if circumstances change, and that the list of approved companies can shift too.

Fable 5 Is Still in Limbo

The letter didn't say a word about Fable 5's status. That model remains offline for now. People close to the discussions say the government is leaning toward releasing Fable 5 as well, but nobody's given a clear timeline for when that might happen.

An Anthropic spokesperson said the company has received the notice and is now working to get the approved companies back online with Mythos 5 as fast as possible.

The spokesperson added that Anthropic is glad to see this progress and plans to keep working with the government, with the goal of eventually making Fable 5 available to everyone again.

A Rocky Few Months for Anthropic

This isn't the first friction point between Anthropic and the federal government this year, and it's playing out as the company prepares for an IPO. Earlier on, Anthropic refused to let the U.S. military use its AI models for domestic surveillance or for fully autonomous weapons systems. The government's response was to place the company on a national security blacklist.

The timing of this latest move is notable too. Lutnick's letter landed the same day that Anthropic's main competitor, OpenAI, rolled out its newest model, GPT-5.6, to its own short list of government-approved partners.

It's a sign that this kind of government gatekeeping over the most powerful AI models might be becoming the new normal across the industry, not just a one-off problem for Anthropic.

There's also more uncertainty on the horizon. A separate cybersecurity executive order has an August deadline attached to it, requiring federal agencies to set up a formal process for evaluating how risky an AI model's cyber capabilities might be.

That means even after this de-escalation, the rules of the road for releasing frontier AI models are still very much being written in real time.