Home / Glossary / AI identity security
Glossary
What is AI identity security?
AI identity security is the practice of knowing which identity each AI agent runs as and limiting what that identity can reach. An agent's identity is the employee, the account it signs in with and the connectors that link it to other systems. AgentTracer is the AIBound product that maps it.
By the AIBound team · Updated 17 November 2026
Key takeaways
- An AI agent's identity is the employee, the account and the connectors it uses to reach other systems.
- An agent on a laptop usually runs with the employee's access, so it can reach what the employee can reach.
- Security has to map what each agent can reach before it can set a rule to limit it.
Why AI agent identity matters
An AI agent rarely has an account of its own. It borrows one. A coding agent on a laptop uses the developer's logins, tokens and keys. If the developer can open the production console, so can the agent.
That makes agents hard to review. An access review lists people and service accounts, and agents appear on neither list. A personal login wired into a company system is harder still to spot. If an agent is tricked or compromised, the attacker gets the same reach the agent had.
How AI identity security works
AI identity security follows the path from a person to the data an agent can touch. Each step on that path is a place to set a limit.
- 1 Find each AI agent and the employee who runs it.
- 2 Identify the account the agent uses, and whether it is a company account or a personal one.
- 3 Map the connectors, tokens and systems that account opens for the agent.
- 4 Cut the access the agent does not need, and set rules for the access it keeps.
AI identity security vs identity and access management
Identity and access management (IAM) controls which people and service accounts may sign in and what they may open. AI identity security looks at the agents that act through those accounts.
| Question | AI identity security | Identity and access management |
|---|---|---|
| Who is the subject? | AI agents acting through an account | People and service accounts |
| What does it read? | The device, the agent, its account and its connectors | The directory, sign-ins and granted permissions |
| What does it show? | What each agent can reach, and on whose account | What each account can reach |
Sources
- OWASP Non-Human Identities Top 10 · owasp.org
- NIST glossary: least privilege · csrc.nist.gov
Sources checked 19 November 2026
Examples
AI agent identity examples
FAQ
Frequently asked questions
What is AI identity security?
AI identity security is the practice of knowing which identity each AI agent runs as and limiting what that identity can reach. It traces each agent from the employee, through the account and connectors it uses, to the systems and data it can touch. Then it sets limits on that access.
What is AI agent identity?
AI agent identity is who or what an AI agent runs as: the employee, the account it signs in with and the connectors that link it to other systems. Most agents on a laptop borrow the employee's own access. That means the agent can reach whatever the employee can reach.
Do AI agents count as non-human identities?
Often, yes. A non-human identity is an account used by software, such as a service account or an API key. An AI agent that has its own credentials fits that label. Many agents on employee devices use a person's account, so they sit between the two groups.
What can a compromised AI agent access?
A compromised AI agent can reach whatever its employee can reach: repositories, cloud consoles, tokens, and HR or finance systems. AgentTracer maps that path, from the person through the agent's account and connectors to the data, so you can see how far an attack could spread before it happens.