The Trump administration's order to pull Anthropic's latest cybersecurity models appears more retaliatory than technical, according to analysts. The vague national security export control shows the AI industry is not immune to government interference.
The U.S. Commerce Department sent Anthropic an export control directive banning non-Americans, including Anthropic employees, from accessing Fable 5 and Mythos 5, citing unspecified national security concerns. Anthropic shut down both models to all customers to comply. The government did not provide specific details, and Anthropic suspects it relates to a guardrail bypass but is unsure.
According to Axios, the directive stemmed from 'personality differences' between Anthropic and the Trump administration rather than a technical issue. Cybersecurity expert Katie Moussouris stated the bypass described in a private paper should never have triggered export controls, as it involved asking the model to 'review code' versus 'fix code'—with nearly identical outcomes.
This incident demonstrates the AI industry's vulnerability to direct government intervention. Moussouris and dozens of security experts called for revoking the order, warning that removing advanced cybersecurity capabilities from defenders is dangerous. Past administrations have similarly overreached, nearly outlawing legitimate security tools with broad export language.