According to people familiar, the Trump administration has decided to block
foreign access to Anthropic’s most capable AI models after Amazon CEO Andy Jassy
told US officials, including Treasury Secretary Bessent, that Amazon researchers
were able via a series of prompts to induce Anthropic’s Fable 5 to produce
information that could be used in cyberattacks. The White House convened,
security researchers corroborated Amazon’s findings, and the government — with
presidential approval — will bar foreign governments, companies and individuals
from accessing the tool. Anthropic said it will cut all user access to Mythos
and Fable to ensure compliance; US officials have long judged Anthropic’s new
models difficult to manage from a security standpoint.
