In the context of agents, identity means two things at once: the agent's defined persona (what role it plays, whose behalf it acts on) and the authentication material it carries — API keys, OAuth tokens, session credentials, delegated user sessions.

The problem: most agents don't have a clean, governed identity of their own. They borrow one — usually from the user who set them up, or the system that deployed them. That borrowed identity is the attack surface.

Privilege refers to what the agent is allowed to do — what systems it can access, what operations it can perform, what data it can read or write. In traditional security, privilege is carefully bounded. In agentic systems, it often isn't.

Agents inherit privilege from their operators, delegate it to sub-agents, and carry it forward across tasks. Each hand-off is an opportunity for it to expand, leak, or persist longer than intended.

Abuse is the key word. This risk isn't about external attackers breaking in — it's about those identity and privilege mechanisms being exploited, bypassed, or manipulated to give someone (or something) more access than they were supposed to have.

Sometimes the abuser is an external attacker. Sometimes it's a misconfigured agent. Sometimes it's a legitimate user who discovered they could escalate their access through an agent that trusted them more than it should.