Anthropic Pulls Claude Fable and Mythos AI Models After Feds Claim Jailbreak

Anthropic has temporarily withdrawn access to its newly released AI models, Mythos 5 and Fable 5, following a directive from the U.S. government. The decision came after federal authorities identified a method to 'jailbreak' Fable 5, allowing users to bypass safety protocols designed to prevent misuse. The government classified the models as subject to export controls, forcing Anthropic to suspend access for all foreign nationals, both inside and outside the U.S. The restrictions apply broadly, requiring Anthropic to shut down the models entirely to comply. The company expressed disagreement with the scope of the action, stating it could stifle innovation in frontier AI models if applied broadly. Anthropic emphasized the need for a transparent, fair, and fact-based process for blocking unsafe deployments, which it claims this action does not meet. In response, Anthropic dispatched staff to Washington to negotiate with the Trump administration, including meetings with Commerce Secretary Howard Lutnick and National Cyber Director Sean Cairncross. The company reportedly spent hours in discussions over the weekend to address the issue. This is not the first time Anthropic has clashed with U.S. authorities over AI restrictions. Earlier this year, the Department of Defense labeled Anthropic a 'supply chain risk' after the company refused to allow its Claude AI models to be used for fully autonomous weapons or mass domestic surveillance. The Pentagon had sought broader access for 'all lawful purposes,' highlighting ongoing tensions between AI developers and government oversight. Anthropic’s statement underscored its commitment to safety but criticized the lack of clarity in the government’s approach. The company has not yet provided further details on whether access to the models will be reinstated or under what conditions. The situation reflects broader challenges in balancing AI innovation with regulatory controls.