Marlene Amstad, who leads the Swiss market regulator FINMA and chairs an international forum on supervisory technology, warned that banks and financial sector supervisors must move faster to adopt new technological defenses as artificial intelligence intensifies cybersecurity risks.
Speaking after an initial hackathon aimed at producing shared tools for market supervisors, Amstad said models that detect software vulnerabilities have recently highlighted a rise in cyberattack and national security risks. Those same developments, she said, are prompting fresh questions about safety and accountability inside financial firms.
"As hackers move faster, banks must adapt by patching vulnerabilities more rapidly," Amstad said in an interview. Her remarks followed a convening of about 100 policy and technology specialists who participated this week in a hackathon designed to jointly build supervisory tools focused on crypto markets.
FINMA has helped establish a forum under the International Organization of Securities Commissions - the market regulation standard setter - with the explicit objective of encouraging adoption of AI by regulators that together oversee roughly 95% of global financial markets. The forum is intended to accelerate capacity building and the creation of shared approaches to supervise AI-driven systems.
Participants at the hackathon worked on prototype tools that supervisors could use when monitoring digital asset markets. Amstad said regulators are exploring options such as embedding safeguards directly into digital asset systems, an approach under consideration as part of wider efforts to reduce operational exposure.
Experience with some recent AI models has underscored these vulnerabilities. Amstad pointed to models such as Anthropic's Mythos as having revealed operational risks tied to AI deployment. Those concerns have already fed into official government action: this month the U.S. government ordered Anthropic to suspend exports of its latest Mythos and Fable AI models on national security grounds.
The broader vendor and industry reaction to those risks is also visible. Chinese cybersecurity firm 360 Security Technology said this week it has developed a domestic response to Mythos, reflecting demand from some markets for locally controlled alternatives.
Amstad emphasized that Switzerland needs to maintain access to the most advanced AI models even as it works to deploy stronger safeguards. She said AI itself will be essential to harden systems prior to deployment and to help supervisors and firms identify and remediate weaknesses more quickly.
The discussion signals a push by supervisors to combine shared technological tools, cross-jurisdictional cooperation, and vendor engagement to address AI-related safety and accountability questions in the financial sector. The initiative is at an early stage - evidenced by the inaugural hackathon - but it reflects a recognition among regulators that new capabilities will be required as cyber threats evolve alongside AI capabilities.