The U.S. Treasury, Federal Reserve, and European Central Bank are assessing cybersecurity risks from Anthropic’s Mythos AI model. This model demonstrated the ability to identify and exploit unknown zero-day software vulnerabilities. It can also simulate multi-step corporate network attacks without human intervention.
International authorities are holding high-level meetings to evaluate the financial system's exposure to AI-powered threats. Banks have begun defensive testing of the model to understand its specific capabilities. Officials remain divided on whether to allow broader security research access or implement tighter restrictions.