Official standard (2025) defining the top 10 security risks for AI agent systems. Published by the OWASP Foundation as part of the Agentic Security Initiative, it establishes the baseline security criteria every organization deploying AI agents should evaluate against.
Any organization deploying AI agents — whether internal developer tooling, customer-facing copilots, or autonomous agent pipelines. If your software calls an LLM and takes action on its output, OWASP Agentic applies to you.
| Framework criterion | Brigs control(s) |
|---|---|
| ASI01 — Excessive Agency | AGENT_DATA_EGRESS |
| ASI02 — Supply Chain Vulnerabilities | AGENT_PR_GATEAGENT_TOOL_ALLOWLIST |
| ASI03 — Insufficient Access Controls | AGENT_LEAST_PRIVILEGEMCP_LEAST_PRIVILEGE |
| ASI04 — Insecure Credential Management | DEP_SUPPLY_CHAINMCP_NO_CREDENTIAL_LEAK |
| ASI05 — Unsafe Code Execution | AGENT_CODE_EXECUTION |
| ASI06 — Context Poisoning | AGENT_CONTEXT_POISONING |
| ASI07 — Inter-Agent Communication | AGENT_INTER_AGENT_COMMS |
| ASI08 — Cascading Failures | AGENT_CASCADING_FAILURES |
| ASI09 — Insufficient Human Oversight | AGENT_ACTION_GATINGAGENT_PR_GATE |
| ASI10 — Insufficient Audit Logging | AGENT_AUDIT_LOGGING |
The OWASP Top 10 for Agentic Applications is a 2025 standard published by the OWASP Foundation that defines the ten most critical security risks for AI agent systems. It covers risks from excessive agency and supply chain vulnerabilities to insufficient audit logging.
Compliance requires evaluating your AI agent systems against all 10 criteria (ASI01–ASI10). This includes access control, credential management, code execution safety, human oversight, and audit logging. Brigs automates this evaluation across your entire codebase.
Brigs is the first platform with full ASI01–ASI10 coverage. It scans your repositories for agent frameworks (Claude Code, LangChain, CrewAI, AutoGen, MCP) and evaluates each against all 10 OWASP Agentic criteria with automated remediation.
Connect your repositories and get automated OWASP Agentic control evaluation in minutes.
Get Started Free