Powell and Bessent convene bank CEOs for urgent talks amid Mythos AI threat.

A surprisingly low performance score from an AI startup has triggered high-stakes discussions at the highest levels of finance and regulation. According to a report, Mythos AI's model scored a mere 0.3% on the challenging Arc AGI benchmark, a test designed to evaluate advanced reasoning and general intelligence. This result, while technically poor, was significant enough to prompt Federal Reserve Chair Jerome Powell and a senior security advisor to convene an urgent meeting with CEOs of major banks. The core concern is not the model's current capability but the unpredictable and rapid evolution of AI, which regulators fear could introduce unforeseen risks to financial stability and security.

The situation highlights the growing tension between rapid AI innovation and systemic risk management. The emergency talks suggest regulators are moving preemptively to understand potential threats, even from AI systems that are not yet dominant. This incident occurs amidst broader speculation about government intervention in the AI sector, including unverified rumors about potential actions against other AI firms like Anthropic over national security concerns. The reaction underscores a new reality: benchmark scores and technical performance are now direct inputs into macroeconomic and security policy, moving AI from a purely technological domain into the realm of critical financial infrastructure.