Federal Reserve Chair Jerome Powell and Treasury Secretary Scott Bessent called an urgent meeting with the CEOs of America’s largest banks this week to discuss cybersecurity risks posed by Anthropic’s new AI model, Mythos. The Information first reported the meeting, which took place at Treasury headquarters in Washington on Tuesday.
The gathering included top executives from Citigroup, Bank of America, Wells Fargo, Morgan Stanley, and Goldman Sachs. JPMorgan’s Jamie Dimon was the only major banking CEO who couldn’t attend. The message from Powell and Bessent was clear: make sure your systems are ready for what’s coming.
What Is Mythos and Why the Alarm?
Anthropic launched Mythos earlier this week but stopped short of a broad public release. The reason? The model demonstrated an extraordinary ability to discover previously unknown cybersecurity vulnerabilities, commonly called zero-day bugs. Reports indicate Mythos uncovered thousands of these flaws across critical software systems.
That’s a double-edged sword. In the right hands, it’s a powerful defensive tool. In the wrong hands, it’s a skeleton key to some of the most sensitive financial infrastructure on the planet. That’s exactly what brought Powell and Bessent to the table.
Project Glasswing: Controlled Deployment
To manage the risk, Anthropic created Project Glasswing, a collaboration with major technology and cybersecurity companies. The initiative uses Mythos to identify and patch vulnerabilities in critical software before attackers can exploit them. Think of it as a controlled burn: use the model’s power to fix problems proactively rather than wait for bad actors to find them first.
This is a significant shift in how AI companies handle powerful capabilities. Rather than releasing the model broadly and hoping for the best, Anthropic is restricting access and partnering with defenders. It’s a model (no pun intended) that other AI labs will likely have to adopt as their systems grow more capable.
Why This Meeting Matters
A few things stand out here:
- Government involvement at the highest level. The Fed Chair and Treasury Secretary don’t typically convene emergency meetings with bank CEOs over a single AI model launch. This signals that policymakers view advanced AI capabilities as a direct threat to financial stability.
- Banks as critical infrastructure. The financial sector is one of the most targeted industries for cyberattacks. A model that can find zero-days at scale changes the threat calculus overnight.
- Precedent for AI governance. This is one of the first times we’ve seen a coordinated government-industry response to a specific AI model’s capabilities, not a hypothetical risk, but a concrete one.
What Comes Next
Expect increased regulatory attention on frontier AI models with cybersecurity implications. Banks will likely need to accelerate their AI security audits and potentially integrate tools like Glasswing into their defensive posture.
For the broader AI industry, the Mythos situation raises a question that isn’t going away: when a model is powerful enough to threaten critical infrastructure, who decides how it gets deployed? Anthropic chose a controlled approach. Not every lab will make the same call.
More details on the meeting and its outcomes are available at The Information’s original reporting.