Amazon security research reportedly led to the White House’s Anthropic Fable ban

The Wall Street Journal reports that Amazon cybersecurity research and CEO Andy Jassy’s discussions with the White House helped trigger export controls affecting Anthropic’s Fable 5 and Mythos 5. Amazon’s paper allegedly found that Fable 5 could be prompted to produce information useful for cyberattacks, leading officials to block access by foreign nationals.

The episode shows how AI safety concerns are increasingly merging with national security and export policy. If model access can be restricted based on adversarial testing, major cloud and AI providers may face growing pressure to prove their systems cannot enable dual-use cyber capabilities.