Global finance leaders flag serious concerns about Mythos AI model

Finance ministers, central bankers, and senior financiers are increasingly focused on the potential risks posed by Anthropic’s Claude Mythos model, amid fears it could expose critical weaknesses in global financial infrastructure.

The model has already prompted high-level discussions and emergency-style meetings after early testing revealed vulnerabilities across major operating systems and widely used applications. Officials and industry experts say the system may have an “unprecedented” ability to detect and exploit cybersecurity flaws, though some caution that its full capabilities are still not fully understood.

Canadian Finance Minister François-Philippe Champagne said the issue dominated conversations during this week’s International Monetary Fund meetings in Washington.

“Certainly it is serious enough to warrant the attention of all the finance ministers,” he said, adding that unlike physical risks, the challenge with AI is “the unknown, unknown.”

He stressed the need for safeguards, saying authorities must ensure “we have process in place to make sure that we ensure the resiliency of our financial systems.”

Major banks and government agencies are now being given early access to Mythos to assess vulnerabilities before any broader rollout.

C. S. Venkatakrishnan, chief executive of Barclays, said the concerns are significant enough to demand immediate attention.

“It’s serious enough that people have to worry,” he said. “We have to understand it better, and we have to understand the vulnerabilities that are being exposed and fix them quickly.”

He added that the situation reflects a more connected financial system where risks and opportunities are increasingly intertwined.

Anthropic has indicated that Mythos has already uncovered multiple flaws across operating systems, financial platforms, and web browsers. In response, access has been restricted to a small group of institutions, including major technology firms and systemically important banks, allowing them to strengthen defenses before wider exposure.

Authorities in the United States have taken similar steps. The Treasury Department has encouraged leading banks to deploy the model internally to identify weaknesses, while also exploring ways to make a controlled version available to federal agencies. A memo from the White House Office of Management and Budget outlined plans to introduce safeguards before any such access is granted.

Andrew Bailey, governor of the Bank of England, said the implications for cybercrime must be taken seriously.

“We are having to look very carefully now what this latest AI development could mean for the risk of cyber crime,” he said, warning that such tools could make it easier for “bad actors” to identify and exploit system vulnerabilities.

Senior US officials, including Scott Bessent and Jerome Powell, have already convened Wall Street executives to address the risks. Attendees reportedly included leaders from major banks such as Goldman Sachs, Bank of America, Citigroup, and Morgan Stanley, underscoring the systemic importance of the issue.

Industry voices suggest the concerns may not be limited to Anthropic. Sources indicate another US AI company could release a similarly capable model without comparable safeguards.

James Wise of Balderton Capital described Mythos as “the first of what will be many more powerful models” capable of exposing system vulnerabilities. His Sovereign AI unit is investing in companies focused on AI security, adding, “We hope the models that expose vulnerabilities are also the models which will fix them.”

Mythos is part of Anthropic’s Claude family of models, a competing system to offerings from OpenAI and Google. Unlike previous releases, the company has restricted access due to concerns that the tool could be misused to uncover sensitive flaws or break into protected systems.

Internal testing raised alarms after the model identified critical bugs that would typically require highly skilled hackers to discover. Some vulnerabilities reportedly date back decades, highlighting gaps that had gone undetected by traditional security systems.

The concerns have also spilled into policy disputes. The Pentagon recently designated Anthropic as a potential supply chain risk, a move usually reserved for foreign adversaries. The company successfully challenged a proposed ban in court, arguing it would result in significant financial losses.

Within national security circles, Mythos has introduced new uncertainty around how cyber threats are assessed. One official described the impact as comparable to equipping an ordinary hacker with tools akin to those used by elite operators.

Despite the risks, authorities continue to engage with Anthropic. Federal agencies are preparing for possible controlled access, while regulators and financial institutions race to understand and address the vulnerabilities the model has already begun to uncover.