Powell and Bessent Summon Bank CEOs Over Anthropic Mythos AI Threat

Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell took the extraordinary step of summoning Wall Street's most powerful executives to an emergency meeting on Tuesday. The topic: Anthropic's new Mythos AI model and the cybersecurity earthquake it represents for the global financial system.

The meeting signals a new chapter in AI governance — one where AI models are powerful enough to warrant direct intervention from the highest levels of government.

What Happened

Bessent and Powell assembled the group at Treasury headquarters in Washington on April 10. The guest list read like a who's-who of American finance: Citigroup CEO Jane Fraser, Morgan Stanley CEO Ted Pick, Bank of America CEO Brian Moynihan, Wells Fargo CEO Charlie Scharf, and Goldman Sachs CEO David Solomon all attended. JPMorgan CEO Jamie Dimon was invited but unable to attend.

The message was blunt — banks need to take Mythos seriously and start deploying its capabilities to find vulnerabilities in their own systems before someone else does.

Why Mythos Has Washington Worried

Mythos is unlike any AI model that's come before it. Designed specifically for cybersecurity, it doesn't just scan code for known patterns — it reasons across entire systems, chaining vulnerabilities into real attack paths that human security teams have missed for decades.

The numbers are staggering. Anthropic's internal testing revealed that Mythos discovered a 27-year-old vulnerability in OpenBSD, an operating system widely considered one of the most secure in existence. The model also identified numerous 10- and 20-year-old vulnerabilities that had evaded human detection.

Why this matters: if Anthropic built a model that can find these flaws, it's only a matter of time before similar capabilities end up in the wrong hands. That's the scenario keeping Bessent and Powell up at night.

Project Glasswing Takes Shape

The meeting came alongside the expansion of Project Glasswing, a joint initiative between Anthropic, JPMorgan Chase, Amazon, Apple, Google, Microsoft, and NVIDIA. The project's mission is straightforward — use Mythos to secure critical infrastructure before adversaries develop comparable tools.

Anthropic has committed $100 million in Mythos Preview usage credits to 40 organizations that build and maintain critical infrastructure. The goal is to give defenders a head start in the AI-powered cybersecurity arms race.

Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley are already testing Mythos internally. More banks are expected to gain access in the coming days.

The Broader Implications

This meeting marks the first time U.S. financial regulators have convened bank CEOs specifically over an AI model's capabilities. It sets a precedent for how governments may respond as AI systems grow more powerful.

The concern extends beyond banking. Any organization running software — and that's every organization — faces the same exposure. Mythos scored 83.1% on CyberGym's vulnerability reproduction benchmark (a standardized test for finding and exploiting software flaws), compared to 66.6% for Claude Opus 4.6. That gap represents the difference between a tool that helps security teams and one that fundamentally changes the threat landscape.

For banks specifically, the stakes are existential. Financial institutions run on legacy software systems, some dating back decades. Those systems contain the exact kind of deeply buried vulnerabilities that Mythos excels at finding.

What Banks Should Do Now

The immediate action is clear: get access to Mythos Preview and start scanning critical systems. Anthropic rolled out Mythos this week in a limited capacity specifically because of dual-use concerns — the same model that finds vulnerabilities for defenders could theoretically be used by attackers.

Beyond individual action, the meeting signals a shift toward coordinated defense. Project Glasswing represents a template for public-private AI security collaboration that could expand well beyond financial services.

The Bottom Line

When the Treasury Secretary and Fed Chair personally summon bank CEOs to discuss an AI model, it's not a drill. Mythos represents a genuine inflection point — the moment AI capabilities crossed the threshold from theoretical risk to concrete national security concern. Financial institutions that move quickly to adopt defensive AI tools will be better positioned. Those that wait may find their vulnerabilities discovered by someone far less friendly than Anthropic.