US government orders Anthropic to disable two AI models abroad

The Trump administration has ordered Anthropic to immediately block access outside the United States to two of its advanced AI models, Fable 5 and Mythos 5, after security concerns that their safeguards could be bypassed. Anthropic said it shut down the models to comply with the government directive.

The move came after US Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei saying the models would fall under export restrictions. The Commerce Department did not disclose the specific security threat, but Anthropic said regulators had identified a method that could bypass Fable 5’s safety systems.

According to the company, the technique relies on reading software identification code and could let users evade parts of the model’s protections. Anthropic argued the capability is not unique to its systems and already exists in other public models, including OpenAI’s GPT-5.5, and is also used by cyber experts for defensive purposes.

The decision adds to tensions between Anthropic and President Donald Trump’s administration. In February, the government moved to restrict the company’s products in federal agencies after Anthropic called for stricter limits on Pentagon use of its technology. The National Security Agency nevertheless continued using Mythos 5 for offensive cyber operations. It remains unclear how the latest action will affect the AI industry, but experts say it could be a major regulatory step with broader sector-wide implications.