US Reverses Export Ban on Anthropic AI Models

The regulatory U-turn arrives after a brief, chaotic period where Anthropic was forced to pull the plug on its models globally because it could not verify the nationalities of its users. The conflict originated in June, just days after the Fable 5 launch, when the government flagged a vulnerability that allegedly allowed users to bypass safety filters. While Anthropic dismissed the flaw as minor, the White House responded by restricting access, a move many observers linked to the company’s earlier refusal to grant the Pentagon unrestricted use of its surveillance technology.

White House adviser Marc Andreessen noted that the administration remains divided between competing strategies: the desire to saturate global markets with American technology to preempt Chinese influence and the urge to lock down powerful models to prevent potential cyberattacks on critical infrastructure. This tension has left AI firms in a state of regulatory uncertainty.

Currently, Anthropic has resumed global access to Fable 5, while Mythos 5 remains restricted to a select group of American cybersecurity partners. The broader international status of the models, including access for entities like the European Commission, remains in limbo. Clarity may emerge in August, when the White House is expected to finalize new security criteria for advanced AI systems. In response to the ongoing scrutiny, Anthropic has pledged to collaborate with industry giants including Amazon, Microsoft, and Google to establish standardized protocols for identifying and mitigating vulnerabilities in high-level AI.