AI agent identity and autonomy risks

As AI agents get their own credentials (non-human identities), attackers can forge fake agents or hijack abandoned 'zombie' ones to impersonate trusted systems. These identity weaknesses are a key stepping stone in autonomous attack chains where AI systems act on their own to cause harm without human oversight.