AI agents are projected to be a significant insider threat in 2026, as highlighted by Palo Alto Networks. With 40% of enterprise applications integrating AI, security teams face pressure to ensure these agents are secure, as they may have broad access to sensitive data. The emergence of AI also creates risks like privilege abuse and “doppelganger” scenarios, where AI mismanagement could lead to unauthorized actions, such as fraudulent transactions on behalf of executives. Attackers can exploit AI systems to automate attacks, enhancing their capabilities significantly. Best practices for limiting AI permissions and monitoring behavior are crucial to mitigate these threats.
https://www.theregister.com/2026/01/04/ai_agents_insider_threats_panw/