The Trump administration has removed the restrictions it placed on Anthropic's most powerful AI models, allowing the company to bring them back online. According to the Associated Press and numerous outlets, the government lifted the controls "after a cybersecurity alarm" that had prompted the crackdown in the first place.
Anthropic says the affected products were its Claude Fable 5 and Mythos 5 models. According to Euronews, citing Anthropic, the U.S. lifted export controls on the powerful models. Wccftech reports the action came from the Commerce Department, which it says never publicly explained the controls it was now reversing.
The return was not unconditional. Wired reports that Anthropic added a new security measure to get back into the administration's good graces, noting "there were strings attached." According to MarkTechPost, Anthropic redeployed Claude Fable 5 on July 1 alongside a new cybersecurity classifier that blocks the technique described in an Amazon report more than 99% of the time, routing flagged requests to its Opus 4.8 model. MarkTechPost also reports the company proposed a four-criteria jailbreak severity framework developed with Amazon.
Business Insider Africa reports the restoration followed negotiations with the White House. The changes have not pleased everyone: PCWorld reports that Claude subscribers are "furious" over Fable's new restrictions, and MakeUseOf cautions users not to get too attached.
Why it matters: The episode shows how directly the U.S. government can now switch advanced AI systems on and off, making national-security concerns and White House negotiations a routine part of which AI tools the public can actually use.