The European Central Bank held a meeting focused on cybersecurity risks posed by new artificial intelligence models and plans to pressure banks to accelerate their system security reinforcement efforts.
Frank Elderson, a member of the ECB Executive Board, stated: "For years, we have been working with banks on various cybersecurity issues, yet problems persist. With the rapid advancement of AI technology, security vulnerabilities must be addressed more swiftly."
Elderson indicated that at Tuesday's meeting, the ECB will brief attendees on potential threats to the financial system identified by AI models such as the Claude Mythos preview. The ECB will also require U.S. banks with access to this technology to share relevant analysis with their European counterparts, who may not be able to participate in testing.
Elderson warned banks to promptly deploy software patches, as AI can identify system vulnerabilities within minutes of a patch's release. He emphasized that European banks cannot use their lack of access to the model as an excuse for inaction, as malicious actors will soon gain access to similar technology.
The restricted access to the Mythos model, which can detect unknown vulnerabilities in information systems, has raised widespread concern within Europe's financial sector. On May 4th, European Commission Executive Vice President Valdis Dombrovskis stated that the EU is in discussions with Anthropic to arrange for companies and banks to undergo vulnerability testing using the model.
Last month, Anthropic reported that the Mythos model had identified thousands of high-risk vulnerabilities in mainstream operating systems and browsers. Meanwhile, the French AI firm Mistral AI is in talks with major European banks to offer its own proprietary vulnerability detection product.
Comments