Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell summoned Citigroup CEO Jane Fraser alongside peers from systemically important banks for a private briefing on risks posed by Anthropic’s new Mythos AI model. The Tuesday session in Washington aimed to ensure that major financial institutions understand the potential attack vectors exposed by powerful AI tools and prepare adequate defenses.

Anthropic’s Mythos model demonstrated the ability to identify thousands of zero-day vulnerabilities across dominant operating systems and browsers, including a 27-year-old flaw in OpenBSD and an undetected vulnerability in FFmpeg. These findings underscore the model’s advanced coding and reasoning capabilities, which can surface security gaps that traditional tools have missed.

For Citigroup, the meeting signals heightened regulatory scrutiny and the likelihood of increased investment in cybersecurity infrastructure and incident response teams. Jane Fraser and her team must now evaluate existing defenses, prioritize patch management, and assess the bank’s exposure to AI-driven threat actors.

Citigroup is expected to explore AI-powered scanning and threat detection platforms, strengthen collaboration with technology partners, and bolster internal training on emerging attack techniques. The bank may also allocate additional budget to cybersecurity standards compliance and continuous vulnerability assessments to mitigate Mythos-style risks.