U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell convened an urgent meeting with top bank executives this week to discuss cybersecurity risks posed by Anthropic's latest AI model, Mythos. The meeting, held at the Treasury Department in Washington, aimed to ensure banks are aware of potential vulnerabilities and are taking steps to defend their systems.
Immediate Action & Core Facts
Anthropic's Mythos model, which can identify and exploit software vulnerabilities, has raised concerns among U.S. officials and financial institutions. The model has reportedly uncovered thousands of previously unknown flaws across major operating systems and web browsers. In response, Anthropic has restricted access to a small group of cybersecurity organizations and launched Project Glasswing, a partnership with major tech companies to use Mythos for defensive purposes.
Deeper Dive & Context
Meeting Details
The meeting included executives from Citigroup, Bank of America, Wells Fargo, Morgan Stanley, and Goldman Sachs. JPMorgan Chase CEO Jamie Dimon was invited but could not attend. The meeting was called while most CEOs were already in Washington for other meetings, indicating the urgency of the situation.
Anthropic's Response
Anthropic has emphasized that Mythos' capabilities emerged from broader improvements in the model's coding, reasoning, and autonomy. The company has stated that the same improvements that make the model effective at patching vulnerabilities also make it effective at exploiting them. To mitigate risks, Anthropic has donated $4 million to open-source security organizations and is not planning to release Mythos to the public.
Global Reactions
While U.S. officials and banks are taking steps to address the risks, analysts in China have noted a more cautious approach, prioritizing financial stability and monitoring potential contagion from the U.S. market. Experts have warned that immediate cybersecurity assessments are needed to avert potential economic damage running into the hundreds of billions of dollars.
Project Glasswing
Anthropic's Project Glasswing involves partnerships with Amazon, Apple, Google, Microsoft, JPMorgan Chase, and others. The goal is to use Mythos to find and patch vulnerabilities before similar AI capabilities spread to attackers. Anthropic has committed up to $100 million in usage credits to support the effort.
Long-Term Implications
The meeting highlights the growing concerns about AI's potential to both enhance and threaten cybersecurity. As AI capabilities advance, the need for robust defensive measures and international cooperation becomes increasingly critical. The Biden Administration has previously identified AI as a potential risk to financial stability, underscoring the importance of proactive measures.