A sudden, urgent summons brought the CEOs of Wall Street’s most powerful banks to Washington D.C. on Tuesday. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell convened the meeting, a stark warning about a new and escalating cybersecurity threat.
The source of their concern? Anthropic, an AI giant, and its latest creation: Claude Mythos Preview. Experts believe this model represents a fundamental leap in artificial intelligence, one with potentially devastating consequences for financial security.
Those present represented institutions critical to the global economy – Goldman Sachs, Citigroup, Morgan Stanley, Bank of America, and Wells Fargo. Each bank is designated “structurally important” by the Federal Reserve, highlighting the gravity of the situation.
The core of the worry lies in Mythos’s unprecedented ability to autonomously discover and exploit weaknesses in software. Anthropic itself claims the AI surpasses nearly all human experts in identifying vulnerabilities, uncovering flaws even in systems considered highly secure.
Thousands of previously unknown software defects have already been identified by Mythos, some dating back decades within established companies. This capability dramatically increases the potential for frequent and destructive cyberattacks, potentially empowering adversaries.
Anthropic has briefed senior U.S. government officials on the risks posed by Mythos, though the specific agencies involved remain undisclosed. The implications are clear: this technology demands immediate attention and robust defenses.
The relationship between Anthropic and the U.S. government is complex and recently fractured. Once a key partner, securing a $200 million Pentagon contract in 2025, the company drew a firm line against its technology being used for autonomous weapons or domestic surveillance.
This refusal led to a dispute with the Department of War, culminating in Secretary Pete Hegseth designating Anthropic as a supply chain risk – effectively barring federal contractors from utilizing its products. The company’s appeal of this designation was recently rejected by a federal appeals court.
The court’s decision was hailed as a victory for military readiness by Acting Attorney General Todd Blanche, who emphasized the need for the military to have full access to Anthropic’s models if integrated into sensitive systems. Control, Blanche asserted, rests with the Commander-in-Chief, not a technology company.
JPMorgan Chase, notably, is a participant in “Project Glasswing,” an initiative leveraging Mythos to defend against future AI-driven threats. This involvement underscores the dual nature of the technology – a potential weapon and a potential shield.
The urgency of the Treasury and Federal Reserve’s meeting signals a new era in cybersecurity, one where the threat isn’t just about known vulnerabilities, but about an AI actively seeking them out, and the race to defend against a constantly evolving, intelligent adversary.
