Cybersecurity：2026年CISO AI风险报告（英文版）

AI identities don’t need a security sign-off to get into your environments. As teams continue to adopt copilots, agents, and AI-driven apps, this new class of identities is entering business-critical systems unchecked. Sometimes through sanctioned projects, sometimes without approval. Once inside, these agents can act independently or even create additional identities, which means they often begin expanding before security teams can define how to control them. Seventy-one percent of organizat