Back to all articles
•Technology

Amazon CEO's Concerns Prompt U.S. Crackdown on Anthropic AI Models

View original source

Recent reports indicate that Amazon CEO Andy Jassy raised concerns about the security of Anthropic's AI models with U.S. government officials, leading to significant repercussions.

  • The Wall Street Journal stated that Jassy communicated potential security threats posed by Anthropic's Claude Fable 5 model, which Amazon researchers allegedly used to access sensitive information possibly exploitable in cyberattacks.
  • As a result, an export control ban was imposed by the government on Anthropic's Fable 5 and Mythos 5 models.
  • Amazon is a substantial investor in Anthropic, having committed $5 billion to the AI company and future cloud spending estimated at $100 billion.
  • David Sacks, formerly Trump's AI czar, revealed that a reliable source informed both Anthropic and U.S. officials about a security 'jailbreak' related to the models.
  • Sacks recounted that government officials requested Anthropic CEO Dario Amodei to rectify the identified security issues or stop deploying the model. However, Amodei did not comply.
  • An Amazon representative mentioned the routine nature of government consultations on security risks while declining to provide specifics about their discussions.