The US Treasury convened an emergency meeting over Anthropic's new AI model with expanded capabilities for finding software vulnerabilities - media

The head of the US Treasury, Scott Bessent, and the Chairman of the Federal Reserve System (FRS), Jerome Powell, held an urgent meeting this week with the heads of the largest US banks. The reason for this was the release of a new AI model from Anthropic, which the developers presented as a solution with broad capabilities for identifying software vulnerabilities. Bloomberg and Financial Times write about this.

The meeting took place on Tuesday, April 7, at the headquarters of the Ministry of Finance in Washington. Executives from Bank of America, Citigroup, Goldman Sachs, Morgan Stanley and Wells Fargo attended.

According to Bloomberg sources, the heads of the Federal Reserve and the Ministry of Finance wanted to discuss with the banking sector possible risks associated with the Claude Mythos Preview model, which was presented by Anthropic on April 6. The new AI model was presented to a limited number of partners, including Amazon, Apple and Microsoft. This is the first time that the company presents its new development as part of a limited launch, FT notes.

The Mythos model is claimed to be a powerful development capable of identifying vulnerabilities in all major operating systems. The company’s press release on the launch separately clarified that the model can also use so-called “zero-day” vulnerabilities at the direction of users.

Anthropic also claims that the model has already found similar vulnerabilities in all major operating systems and web browsers, some of which are ten or twenty years old. The company explained the limited release by the fact that many of the errors identified by the model have not yet been corrected.

At a meeting between the Ministry of Finance and banks, the possibility of using this and other AI models by attackers was discussed. Specific points made at the meeting are not given.

Bloomberg notes that such a meeting, convened on short notice, suggests that regulators consider new types of cyberattacks using AI to be one of the biggest risks for the financial industry. All banks invited to the meeting are systemically important, which means their stability is a priority for the financial system, the agency emphasizes.

Anthropic already has a difficult relationship with the current US presidential administration. In early March of this year, the US Department of Defense recognized the company as a supply chain threat. The Pentagon's decision was the result of a long-running conflict over the terms of the use of AI for military purposes: the head of the department, Pete Hegseth, demanded that the company allow the use of its AI for everything - including possible surveillance of citizens or the use of autonomous lethal weapons without human intervention. Anthropic waived these terms.

US President Donald Trump then joined the conflict. He declared that all government agencies should stop using Anthropic tools, and also wrote that "The States will not allow a radical leftist, 'progressive' company to dictate how our great army fights and wins wars."

After this, the company received the status of “supply chain risk”, which is usually assigned to companies associated with foreign adversaries. It affects how Anthropic can do business with companies working with the US defense establishment. The developer, in turn, filed a lawsuit against this decision .