Anthropic has taken its two newest AI models offline following an export control order issued by the Trump administration.
The shutdown
The government cited national security concerns in a letter sent last Friday. No details have been released, but the instruction was clear: the models could not be used by foreign nationals. Because Anthropic employs many international staff, the company pulled the software entirely rather than risk a breach.
The affected tools were Fable 5 and Mythos 5. Fable 5 was intended for public use, while Mythos 5 remained restricted to existing users.
Reports indicate the White House received information from Amazon researchers who allegedly found a way to bypass Fable 5’s guardrails. Amazon CEO Andy Jassy raised these concerns with the administration, and the situation escalated quickly over a Friday afternoon and weekend.
Why Anthropic?
Sean O’Kane noted that Anthropic has not maintained a good relationship with the current administration, distinguishing it from other leading AI labs. He suggested rivals might not face a similar crackdown.
However, Rebecca Bellan pointed out that leading cybersecurity experts have signed an open letter asking Trump to revoke the order. They argue it is dangerous to remove advanced cybersecurity capabilities from network defenders in the United States. Anthropic itself stated that similar jailbreaks could potentially be found in several other AI models.
Anthony Ha added that independent security analysis suggests the actual risk from Anthropic is not unique. He believes the conflict stems from a lack of communication between the two sides, causing any risks to be blown out of proportion.
He warned that while other companies might feel advantaged by the move, creating a regulatory environment where they must constantly worry about offending the government is not a healthy landscape.
What it means
There is a cynical view that the pause allows competitors to catch up to where Anthropic was. Yet, the reaction has also been one of curiosity. Previous stories on Ramp analysis showed that major disputes between Anthropic and the Trump administration previously boosted the company. Downloads of Claude increased, with some users viewing it as a more responsible alternative to ChatGPT.
Rebecca Bellan observed that everyone loves a bad boy. The narrative that the models are so powerful they cannot be released to the public may drive demand, with users eager to access what Trump himself has flagged.
