Overview
- The IMF, which published the warning Sunday, said AI tools can uncover software flaws in minutes and could turn cyber risk into a macrofinancial shock by disrupting payments and trust in banks.
- The fund cited Anthropic’s Claude Mythos as an example and warned of “correlated failures,” where shared bugs in common operating systems or browsers let attackers strike many institutions at once.
- Access to protection is uneven, with firms like JPMorgan, Amazon and Microsoft getting early security fixes while many banks in emerging economies lack comparable tools and resources.
- OpenAI acknowledged that agent features in ChatGPT face stubborn threats such as hidden prompt injections that can hijack behavior, and the UK’s cyber agency advised focusing on mitigation rather than full elimination.
- Venture money has swung hard into AI, reaching 61% of global VC funding in 2025, signaling continued scale in finance even as governance gaps and defensive shortfalls remain.