Amazon CEO Andy Jassy informed U.S. Treasury Secretary Scott Bessent that Amazon researchers bypassed security safeguards on Anthropic’s Fable 5 model. The researchers successfully prompted the advanced AI to provide information that could potentially aid cyberattacks.
This disclosure led the Trump administration to ban foreign access to Anthropic’s Fable 5 and Mythos 5 models. Anthropic subsequently disabled both models globally just days after their initial release.
Amazon maintains a substantial investment in Anthropic despite the security warning. This development underscores the complex dynamics between competition and security within the artificial intelligence landscape.