What is agentic AI?
Agentic AI is software in which a language model decides which steps to take toward a goal and carries them out by calling tools such as APIs, databases, browsers and code execution, with limited human review of each step. An AI agent is one such program, and an agentic system can combine several.
Also called AI agents, autonomous agents. Last reviewed .
Why it matters
The security question changes when output becomes action. OWASP's Agentic AI: Threats and Mitigations describes the threats that come with planning, memory and tool use, and the OWASP Top 10 for Agentic Applications for 2026 ranks them. NIST has opened an AI Agent Standards Initiative to develop standards for agents.
For a security team, the practical difference is authority. An agent holds credentials and uses them without a person approving each call, so the controls built around human sign-in, review and approval do not cover what it does.
Related terms
- AI agent security
- AI agent security is the discipline that protects agentic systems.
- Model Context Protocol (MCP)
- The Model Context Protocol is a common way agents reach their tools.
- Excessive agency
- Excessive agency is the risk of giving an agent more tools, permissions or autonomy than its task needs.
- Human of Record
- A Human of Record is the person who answers for what an agent does.
How Guard0 handles it
Guard0 covers the three ways agents arrive in a company: the AI tools people run on their laptops, the agents inside business platforms, and the agents your teams build. Each agent is listed with a named owner and a record of what it did.
Read the guide: The Complete Guide to Agentic AI Security in 2026
Sources
- Agentic AI: Threats and Mitigations. OWASP Gen AI Security Project.
- OWASP Top 10 for Agentic Applications for 2026. OWASP Gen AI Security Project.
- AI Agent Standards Initiative. NIST.