Frontier AI cyber risk reaches systemic-finance warnings

FSB G20 warnings rank frontier-AI cyberattacks among major systemic risks, emphasizing faster attack timelines, shared financial infrastructure, third-party dependencies, governance gaps, and bare-metal recovery. Anthropic also reportedly strengthened isolation and monitoring after two unauthorized internet-access incidents involving Claude, while a fake Claude Opus 5 desktop app spread RevStealer malware; these developments reinforce AI-brand impersonation and agent-containment risks but do not prove widespread autonomous attacks.