Amid Anthropic Mythos fears, IMF chief warns global financial system lagging on AI risks

IMF Managing Director Kristalina Georgieva warns global monetary system at risk from AI cyber threats

The world may be moving fast on artificial intelligence, but its financial safety nets are not keeping up, according to International Monetary Fund Managing Director Kristalina Georgieva.

Speaking on CBS News' "Face the Nation" on Sunday, Georgieva said global systems are not yet equipped to handle large-scale cyber threats linked to AI. Her warning came just before the International Monetary Fund and World Bank begin their spring meetings in Washington.

Georgieva said the international community needs to focus more on building safeguards that can protect financial stability as AI capabilities evolve. She pointed out that while the issue is being discussed in the United States, the risks are not limited to one region.

She called for countries to work together, warning that without coordination, the same vulnerabilities could surface across global markets.

The remarks follow an announcement by Anthropic on April 7, where the company said it would restrict the release of its new "Mythos" model. The decision was linked to concerns over the model's ability to quickly identify security weaknesses.

Earlier this week, Federal Reserve Chairman Jerome Powell and Treasury Secretary Scott Bessent met leading U.S. bank chiefs in Washington, D.C., to discuss cybersecurity risks linked to Anthropic's new AI model, Mythos. The meeting, called on short notice during a Financial Services Forum gathering, brought together top executives from major banks.

The high-level engagement signalled that advanced AI systems are now being viewed as a potential risk to the foundations of the financial system.

The concerns centre on Anthropic's "Claude Mythos Preview," which the company released in a limited capacity, citing fears that its ability to rapidly detect vulnerabilities could be misused by hackers.

The urgency around AI risks has also played out in markets and policy. Reports of the model's capabilities contributed to a decline in cybersecurity stocks last month, while past incidents have shown that AI tools can be used to automate cyberattacks.