White House Blocks Anthropic AI Model Over Cybersecurity Concerns
The Trump administration moved swiftly to block foreign access to Anthropic’s advanced AI model, Fable 5, after Amazon CEO Andy Jassy raised cybersecurity concerns. The decision came after Amazon researchers discovered vulnerabilities in the Mythos-class model that could allow unauthorized access to restricted information, including details on cyberattacks.
While it remains unclear whether Amazon’s tests were conducted at the White House’s request or independently, the Commerce Department invoked national security export controls to halt the distribution of Fable 5 and its underlying model, Mythos 5, to foreign nationals. This includes individuals outside the U.S. as well as non-citizens working domestically, effectively forcing Anthropic to disable the models for all users.
Anthropic prohibits access to its products from within China, yet concerns persisted about potential unauthorized access by Chinese-linked groups.
The unprecedented use of export controls has drawn criticism from AI experts and policymakers, who warn it could lead to a fragmentation of global technology development. Dean Ball, an AI policy expert and former Trump administration official, questioned whether the move targeted Anthropic specifically or reflected broader national security priorities. Meanwhile, Anthropic CEO Dario Amodei reportedly argued that the identified security bypass was limited and did not constitute a full jailbreak of the model’s safeguards.
The decision underscores growing government scrutiny over powerful AI models and could set a precedent for future oversight of AI releases. As senior Anthropic technical staff meet with White House officials, the debate over sovereign AI continues to intensify, with critics warning that such restrictions may deter transparency in AI development.