The Trump administration on Tuesday June 30 lifted all restrictions weighing on Anthropic’s two most powerful artificial intelligence (AI) models, confirming Washington’s regaining control over the activities of this strategic sector, in the name of national security.
“We have received notification that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5,” Anthropic announced on X, specifying that it wanted to “start restoring access” on Wednesday.
We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5.
We’ll begin restoring access tomorrow, and will share an update soon.
We’re grateful to our users for their patience, and to everyone who worked with us on…
— Anthropic (@AnthropicAI) June 30, 2026
I manage my choices I authorize
According to a letter from Commerce Secretary Howard Lutnick made public, the San Francisco mega-startup is committed to “proactively detecting and addressing security risks” related to its models, to “working diligently with the US government on the protocols, standards and outputs of Mythos, Fable and its future models”, and to inform it of “any malicious activity”. The government, specifies the letter, reserves the right to reinstate the restrictions if the company does not honor these commitments.
On June 12, Washington brutally forced Anthropic to cut off access to these two cutting-edge models, citing national security after the detection of flaws likely to cause the full power of these tools to fall into the wrong hands.
On Tuesday, CIA Director John Ratcliffe compared the capabilities of the most advanced AI to “digital nuclear weapons” during a rare public speech in Washington. This forced withdrawal by the government, a first, sparked strong international criticism, reviving debates on digital sovereignty in countries dependent on American technologies.
Anthropic, which has had stormy relations with the Trump administration for months, obtained a first breakthrough on Friday with the unlocking of Mythos 5 for a handful of “cyber defenders and infrastructure operators”, but only American ones. Its foreign partners, notably state cybersecurity agencies in Europe and Asia, remained deprived of it at this stage. Anthropic has not yet clarified whether they are authorized again.
Tuesday’s decision also allows the return online on Wednesday of Fable 5, the general public version of Mythos, restricted on cybersecurity and the risks of biological and chemical attack.
“Voluntary” framework
The affair illustrates the turn taken over several months by the Trump administration. At the beginning of June, a decree established the possibility of federal review of advanced AI models before their commercialization, ensuring that it was not a question of creating a government licensing or permit regime.
However, this is what is emerging in reality, a month later: in addition to the blocking then conditional unblocking of Anthropic, its rival OpenAI agreed to launch its latest model, GPT-5.6, on Friday, with limited access, with government validation client by client. “This is not exactly the process that we consider optimal,” commented OpenAI boss Sam Altman, while judging that the executive “overall is doing a good job in a very difficult situation”.
The chief of staff of the White House, Susie Wiles, for her part praised on
For many observers, the evolution of Washington is likely to favor open and inexpensive models, such as the Chinese DeepSeek, more attractive for clients keen to avoid any dependence on Washington.
Engaged in an unprecedented race, the American tech giants are sinking colossal sums into the construction of data centers and the development of AI models, without having yet proven the profitability of their tools. Valued at nearly a trillion dollars after five years of existence, Anthropic filed its IPO project in June, just like OpenAI.
