Just two weeks ago we had how the United States government blocked two of the most intelligent AI models in the world: Claude Fable 5 and Mythos 5 while Anthropic promised to be working on it to solve it. Just 15 days later, the situation has taken a first turn: the US Department of Commerce has authorized the reactivation of the Claude Mythos 5 model, but not for everyone: only for a closed list of US organizations considered “trustworthy.”
It is an unlock, but with conditions. For now, no Fable 5 newsthe most general and closest model for the average user.
what’s happening. As collects exclusively SemaforHoward Lutnick, Secretary of Commerce in the Trump administration, has sent a formal letter to Anthropic with the good news. The partial redistribution of Mythos 5 will reach a set of more than 100 US organizations that operate and defend critical infrastructure approved by the US government.
Lutnick reported in its letter that Anthropic “has committed to working with the United States government on protocols, standards, and future releases” of its models. Or what is the same, that Anthropic has gone through the government hoop, which now has a voice to choose how and to whom its most advanced models reach. In return, Mythos 5 is partially back.
Why is it important. Because never before has an AI model been subjected to such scrutiny and government approval requirement to be distributed, which establishes a dangerous precedent: converting the most advanced AI models as export technology subject to control, such as semiconductors.
And it doesn’t just affect Anthropic. OpenAI has also launched in the last few hours versions of its new GPT-5.6 model with government-controlled restrictions, such as Sam Altman tells Xwhere despite expressing his predilection for a widespread launch, he shows his willingness to collaborate with the US administration in the authorization of future models.
Context. The June 12 blockade occurred after a warning from Amazon of possible manipulation of Anthropic models for malicious purposes, although reports also weighed that indicated upon his hidden arrival in China. Before the ban, Mythos 5 was available on a limited basis to a number of organizations, among them the government of Spain. Following the blockage, Anthropic sent a team of professionals to work closely with the Department of Commerce and the Office of the National Cybersecurity Director to find a solution.
In detail. Lutnick’s letter specifies that authorized organizations (the list is not public) can use the model in a kind of “white list”: whoever is part of it has access, whether foreigner or not. Curiously, all this is happening without the US having formal regulations to evaluate AI models, unlike the European AI law: The Trump administration blocked first and is building the rules as it goes.
This partial unlocking does not solve the problem for those outside the United States: the list of authorized organizations is exclusively American and the exemption mechanism is based on export control, so any foreign entity wanting to use Mythos 5 would require a specific license that does not exist. So other governments, non-US companies, and foreign consumers still don’t know if or when they will recover these models because the United States is acting unilaterally.
Yes, but. Obviously the unlocking is good news for Anthropic, but it leaves several fronts open. The first is what happens with Fable 5, the general use model, for which there is no return date. The second is that the list of authorized organizations is not public and it is not known what criteria have been used for the selection. In Europe several voices They have already expressed their frustrationalthough the problem of the old continent is another: technological dependence on third parties and the lack of an ecosystem of alternatives to match.
In the background lies another problem that was revealed with the blockade two weeks ago and that the launch of GPT-5.6 has confirmed: the most cutting-edge AI companies and models in the United States have to go through the government filter and we do not know if this will be the modus operandi from now on. If so, this way of proceeding could be a liability in the AI race precisely for the country that currently has the advantage.
In Xataka | China has two ideas to win the AI race: invest a fortune and leave NVIDIA with almost no margin
Cover | Xataka with Gemini
GIPHY App Key not set. Please check settings