U.S. Treasury Secretary Bessent and Federal Reserve Chairman Powell convened an emergency meeting of Wall Street executives on the 7th to discuss the safety risks of the model. On the day that the American AI company Anthropic released its latest AI model, U.S. Treasury Secretary Bessant and Federal Reserve Chairman Powell convened an emergency meeting of Wall Street executives to discuss the security risks of the model.
The meeting, held on the 7th local time at the Treasury headquarters in Washington, was aimed at ensuring that banks are aware of the potential risks of Anthropic’s latest model, Mythos, and other similar models, and take preventive measures to protect the banking system.
It is reported that Wall Street executives summoned to participate in the meeting include Citigroup CEO Jane Fraser, Morgan Stanley CEO Ted Peak, Bank of America CEO Brian Moynihan, Wells Fargo CEO Charlie Scharf, and Goldman Sachs Group CEO David Solomon. People familiar with the matter also said JPMorgan CEO Jamie Dimon was unable to attend.
As of now, all parties participating in the meeting have not commented on the matter.
What are Bessant and Powell worried about?
What makes the U.S. government so nervous is Anthropic’s current strongest model in the field of coding and reasoning, “Claude Mythos Preview” (hereinafter referred to as “Mythos”). This model, which was only officially announced on the 7th, is not yet open to the public and is only available to specific partners.
Although Mythos is a general-purpose model and not specifically designed for cybersecurity purposes, it has been proven to easily identify vulnerabilities in major operating systems and web browsers. Anthropic previously revealed that the model had discovered a 27-year-old vulnerability in OpenBSD, a Unix-like system known for its security and rigor that has now been fixed.
The power of Mythos also means that it can be exploited by hackers to launch cyber attacks. To this end, Anthropic launched a project called "Project Glasswing" (Glass Wings) to provide access to the model to a small number of companies to detect vulnerabilities in their respective products that can be hacked and develop defense mechanisms as early as possible. The first batch of partners of the "Wings of Glass" project include 12 companies including Amazon Web Services, Google, and Microsoft, among which JPMorgan Chase is the only financial institution. Anthropic has now extended model access to more than 40 organizations, but did not provide a specific list.
Running Anthropic once again makes regulators nervous
The emergency meeting also means that U.S. regulators view the possibility of new cyberattacks as one of the biggest risks facing the financial industry.
Anthropic said it had been communicating with the U.S. government about the model's "offensive and defensive cyber capabilities" before launching Mythos. The company also said it is ready to work with local, state and federal representatives in the United States to assist in assessing and mitigating the national security risks associated with artificial intelligence models.
Currently, Anthropic’s lawsuit with the U.S. government is ongoing. The U.S. Department of Defense announced early last month that it had officially designated Anthropic as a "supply chain risk" target, prohibiting the Department of Defense and defense contractors from using the company's artificial intelligence technology in related projects. The company subsequently sued the U.S. federal government, claiming that due to differences between the two parties over the use of the company's products, the other party retaliated and was blacklisted. On the 8th of this month, a U.S. federal court rejected Anthropic’s request to temporarily prevent the Department of Defense from blacklisting it.
Just a few days ago, Anthropic announced that the company's annualized revenue (ARR) has exceeded $30 billion, a significant increase from $9 billion by the end of 2025. The company plans to IPO as soon as the fourth quarter of this year.