IMF, World Bank: AI Models Like Anthropic's Claude Threaten Banking Cyber Defences
International financial officials at the IMF and World Bank meetings this week warned that Anthropic's Claude Mythos Preview AI model poses cyber security threats to the global banking system. Bank of England Governor Andrew Bailey urged regulators to quickly evaluate the risks, while JPMorgan and other banks are testing the model. Regulators are struggling to coordinate a response amid geopolitical tensions.
Senior international financial officials warned that new AI models from US tech companies could threaten the world banking system by exposing cyber defence weaknesses. The concern dominated discussions at the IMF and World Bank spring meetings in Washington this week. Bank of England Governor Andrew Bailey, chair of the Financial Stability Board, stated global regulators must rapidly evaluate the potential cyber security threat from Anthropic's new Claude Mythos Preview model.
Dan Katz of the IMF said the evolution of digital technology poses immense cyber security risks, making it an essential item on the international agenda. Christine Lagarde of the ECB noted Mythos could be "really bad" if it falls into the wrong hands. Anthropic warned Mythos had found thousands of high-severity vulnerabilities and that its capabilities would soon proliferate, potentially beyond safe actors, with severe consequences for economies, public safety, and national security.
Anthropic has shared Mythos with about 40 companies, including Amazon, Apple, and JPMorgan Chase, for testing. Scott Bessent convened leaders of Wall Street banks and Federal Reserve Chair Jay Powell to discuss the weaknesses revealed by Mythos. European regulators and executives say they have been left in the dark about the potential vulnerabilities. OpenAI announced it was sharing its latest model, GPT-5.4-Cyber, designed for cyber security, with vetted security vendors.
JPMorgan CEO Jamie Dimon said Mythos "shows a lot of vulnerabilities that need to be fixed." Morgan Stanley boss Ted Pick said they are working with Claude Mythos and anticipate productivity gains. BNY and Citigroup are also testing Anthropic's model. Regulators are skeptical of a coordinated global response due to geopolitical tensions and are reluctant to hamper AI development.