Global Financial Leaders Warn Advanced AI Could Expose Banking System to Cyber Threats

Senior financial officials from major economies are raising alarms that the latest generation of artificial intelligence (AI) models could pose significant risks to the global banking system by exposing vulnerabilities in cybersecurity defenses.

Some warnings emerged during the International Monetary Fund and World Bank Spring Meetings from April 13 to April 18.

Finance ministers, central bankers, and regulators gathered in Washington to assess mounting economic and geopolitical challenges.

At the center of the discussions was a powerful new AI model, Claude Mythos Preview, developed by San Francisco-based startup Anthropic.

In an April 7 statement, the company said its Claude Mythos Preview model had identified “thousands of zero-day vulnerabilities … many of them critical” across major operating systems and web browsers.

The company emphasized that these findings demonstrate the dual-use nature of advanced AI, which can strengthen defenses while also potentially enabling malicious actors.

Bank of England governor Andrew Bailey, speaking on April 14 at Columbia University in New York City, described the rapid emergence of advanced AI capabilities as a turning point for financial risk.

“It would be reasonable to think that the events in the Gulf are the most recent challenge to us in this world, until, I think it was last Friday, you wake up to find that Anthropic may have found a way to crack the whole cyber risk world open,” Bailey said.

Bailey stressed that regulators must urgently assess what the technology means in practice.

“The issue is: to what extent is this new version of the product going to be able to, in a sense, identify vulnerabilities in other systems which can be exploited for cyberattack purposes,” he said.

Calling the issue persistent and evolving, Bailey added, “It’s the one that never goes away. You have to keep mitigating it, but the threat actors will move on, so we have to deal with it.”

New Concerns

In an April 16 statement, Anthropic said it was proceeding cautiously with the deployment of Mythos.

“We stated that we would keep Claude Mythos Preview’s release limited and test new cyber safeguards on less capable models first,” the company said, adding that its Opus 4.7 model includes safeguards designed to “detect and block requests that indicate prohibited or high-risk cybersecurity uses.”

Anthropic co-founder Jack Clark said on April 13 that the company has been in discussions with the Trump administration about the model, despite the Pentagon’s designation of the company as a supply-chain risk.

Clark characterized the issue as “a narrow contracting dispute,” while emphasizing the firm’s intent to keep government officials informed for national security reasons.

Treasury Secretary Scott Bessent has also convened major banks and Federal Reserve Chair Jerome Powell to discuss the issue.

“Some banks are doing a better job in cybersecurity than others and we wanted to have the ability to convene them and talk about what is best practices and where they should be heading,” Bessent said on April 15.

Regulation, Economic Implications

Across the Atlantic, central banks and regulators are already testing how such technologies could affect financial stability.

The Bank of England said in a letter published on April 16 that it is conducting scenario analysis and simulations to evaluate risks posed by AI. The central bank pushed back against criticism from Parliament’s Treasury Committee that it was taking a passive approach.

Still, concerns remain about regulatory gaps. UK Treasury Committee Chair Meg Hillier said lawmakers were frustrated by delays in applying oversight tools.

Bailey, for his part, argued that financial stability policy needs stronger institutional backing.

On April 14, he said central bank independence in this area is “not robust enough,” and called for a unified framework that treats monetary and financial stability as interconnected.

“I see merit in creating a single overarching narrative with a strong focus on the value of money,” Bailey said, arguing that such an approach would strengthen trust in financial systems.

While cybersecurity risks dominate immediate concerns, officials are also grappling with AI’s long-term economic implications.

European Central Bank President Christine Lagarde, speaking on April 14 at the IMF–World Bank meetings, described two parallel lines of thinking.

“What risks are we opening ourselves to with the artificial intelligence that is currently being developed?” she asked, pointing to the need for rigorous testing and safeguards.

Lagarde added that recent developments have “alerted all of us about the latest software development and the risk that can be identified, for good or the risk that can be identified for bad.”

She also highlighted uncertainty among economists about AI’s broader effects on productivity, employment, and social stability. While some technology firms predict transformative benefits, she said that such outcomes depend on global cooperation.

Fragmentation in data and financial systems, she said, could limit AI’s potential.

European Banking Authority Chair François-Louis Michaud said on April 16 that banks remain “resilient enough” to withstand current shocks, citing strong capital and liquidity buffers.

However, he warned that future risks may differ significantly from past crises.

Michaud added that cybersecurity threats linked to new technologies are “front and center” in regulatory discussions.

Reuters contributed to this report.