U.S. Treasury Secretary Scott Bessent convened a high-level meeting in Washington this week with the heads of the nation’s largest financial institutions to address the systemic cyber risks posed by Anthropic’s latest artificial intelligence model. The gathering, which included Federal Reserve Chair Jerome Powell, focused on the potential for the fresh “Claude Mythos” model to be weaponized by bad actors to destabilize the global financial system.
The urgency of the summons stems from Anthropic’s own admissions regarding the capabilities of Mythos. The AI startup has warned that the model possesses an unprecedented ability to identify and exploit software vulnerabilities, effectively surpassing all but the most elite human hackers. For the U.S. Government, the primary concern is that such a tool could be used to crack encryption or bypass security protocols at “systemically important” banks—institutions whose failure or operational collapse could trigger a wider economic crisis.
The meeting took place even as several bank executives were already in the capital for a lobby group gathering, allowing the Treasury to quickly assemble a group of the most influential figures in American finance. While the Federal Reserve and the Treasury have not issued formal statements regarding the meeting’s specific outcomes, the guest list underscores the gravity of the perceived threat to national security and financial stability.
The ‘Mythos’ Threat: AI vs. Legacy Software
The catalyst for this government intervention is the Claude Mythos model. Unlike previous iterations of AI, Mythos has demonstrated a frightening proficiency in “bug hunting.” According to Anthropic, the model has already exposed thousands of vulnerabilities across various popular applications and software suites. Most alarming is the discovery of flaws that have remained undetected for decades; the company noted that some of the vulnerabilities uncovered by Mythos were up to 27 years old, having evaded both the original creators and professional security monitors.
This capability has forced Anthropic to seize the unprecedented step of restricting the release of the product. In a departure from its usual deployment strategy, the company has limited access to a slight group of partners, including tech giants Microsoft, Apple, and Amazon, as well as networking firms Cisco, and Broadcom. The Linux Foundation has as well been granted access to ensure the security of the open-source operating system that powers much of the world’s server infrastructure.
The risk is not merely theoretical. A recent leak of Claude’s internal code prompted Anthropic to publish a warning early this month, stating that AI models have now reached a point where they can find and exploit vulnerabilities with a speed and precision that rivals the world’s most skilled humans. The company warned that the resulting fallout for public safety, national security, and global economies could be “severe.”
Financial Stability and the ‘Systemic’ Risk
For the Treasury and the Federal Reserve, the intersection of AI-driven cyberattacks and the banking sector represents a “black swan” risk. The banks summoned to the meeting are those classified as systemically important, meaning their operational integrity is viewed as a pillar of the U.S. Economy. The potential for an AI to automate the process of cracking passwords or decrypting sensitive financial data could render traditional cybersecurity defenses obsolete overnight.
The attendee list read like a directory of the American financial establishment:
- David Solomon, CEO of Goldman Sachs
- Brian Moynihan, CEO of Bank of America
- Jane Fraser, CEO of Citigroup
- Ted Pick, CEO of Morgan Stanley
- Charlie Scharf, CEO of Wells Fargo
Notably, JP Morgan Chase CEO Jamie Dimon was invited but unable to attend. Although, Dimon has been vocal about the looming threat. In a recent annual letter to shareholders, Dimon warned that cybersecurity remains one of the firm’s most significant risks and explicitly stated that “AI will almost surely make this risk worse.”
Timeline of Escalation
| Event | Detail |
|---|---|
| March 2026 | U.S. Government designates Anthropic as a supply chain risk. |
| Early April 2026 | Leak of Claude code leads to Anthropic warning about AI vulnerability exploitation. |
| April 2026 | Anthropic restricts “Mythos” release to a select few corporate partners. |
| Current Week | Treasury Secretary summons “systemically important” bank CEOs to Washington. |
Legal Battles and Regulatory Friction
The meeting occurs against a backdrop of escalating tension between the U.S. Government and Anthropic. Only weeks ago, the government designated the AI startup as a supply chain risk—a move that suggests the state views the company’s technology and operational dependencies as potential points of failure for national security. Anthropic is currently fighting this designation in court, arguing against the government’s characterization of its risks.
This legal friction creates a complex dynamic: the government is simultaneously treating Anthropic as a security threat while relying on the company’s internal data to warn the banking sector about the very risks the model creates. The restriction of the Mythos model to a “small clutch” of businesses suggests a recognition by the AI industry that the technology has outpaced the ability of the general public—and potential attackers—to defend against it.
The broader implication is a shift in the cybersecurity paradigm. For decades, the “arms race” was between human hackers and human defenders. With the advent of models like Mythos, the race is now between AI-driven offense and AI-driven defense. If the “offense” (the ability to find flaws) moves faster than the “defense” (the ability to patch them), the structural integrity of the digital financial system could be compromised.
Disclaimer: This article is intended for informational purposes only and does not constitute financial, legal, or investment advice.
The next critical checkpoint will be the ongoing court proceedings regarding Anthropic’s “supply chain risk” designation, which may reveal more about the specific vulnerabilities the U.S. Government is concerned about. Any further updates from the Treasury regarding new cybersecurity mandates for systemically important banks will be closely watched by markets.
We invite you to share your thoughts on the intersection of AI and national security in the comments below or share this story with your professional network.
Related reading
