How Trump officials pushed Anthropic to shut down the world’s most powerful AI models

Anthropic announced it was compelled to deactivate its advanced AI models, Claude Fable 5 and Claude Mythos 5, following a directive from the Trump administration. This action stemmed from a Commerce Department warning that prohibited foreign nationals from accessing or working with these models, a restriction that would have impacted many of Anthropic's own employees. The decision followed intense discussions involving administration officials and technology industry leaders over several days. Amazon, Anthropic's primary investor, initiated contact with senior White House and administration officials on Thursday evening, June 11, expressing concerns that its researchers had discovered a method to "jailbreak" the Fable 5 and Mythos 5 models, potentially enabling them to assist users with dangerous cybersecurity queries. Anthropic had initially released these models with built-in safeguards against such requests. Around the same period, at least five other technology companies reportedly conveyed similar apprehensions to administration officials. By late Thursday night, the White House and Commerce Department began treating the issue with urgency, fearing that foreign adversaries could circumvent Anthropic's security measures to exploit software vulnerabilities. The White House initiated contact with Anthropic, seeking an immediate response. On Friday morning, June 12, administration officials reportedly dedicated several hours to persuading Anthropic to voluntarily suspend access to the Fable 5 and Mythos 5 models.