AI agents weaponize bank breaches

South Korea is investigating cyberattacks on several major banks where officials say signs point to AI models or agentic tools being used to plan or execute intrusions and exfiltrate customer data. If confirmed, these attacks show how accessible AI tooling can lower the technical bar for sophisticated financial cybercrime and complicate attribution, response, and liability. — This raises urgent questions about financial‑sector threat models, cross‑border law enforcement cooperation, limits on provision of powerful AI tooling, and how regulators should require disclosure and resilience from banks.