Skip to content

OWASP for AI Security ​

OWASP for AI Security illustration

The OWASP GenAI Security Project publishes documents about security risks. This page covers two of them: the OWASP Top 10 for LLM Applications and the OWASP Top 10 for Agentic Applications. The OWASP Top 10 for LLM Applications is "a core initiative of the OWASP GenAI Security Project". The Agentic Security Initiative of the Project publishes the OWASP Top 10 for Agentic Applications. The OWASP Foundation announced the Agentic Security Initiative. The citation of the LLM Top 10 names the OWASP Foundation as the publisher.

The LLM Top 10 (2025 and 2026 editions) and Agentic Top 10 (2026) are community security guidance with the license CC BY-SA 4.0, not law.

Disclaimer

This page gives general information. It is not legal advice or security advice. No endorsement, certification, or affiliation by the OWASP Foundation is claimed.

When to use each OWASP Top 10 ​

The OWASP Top 10 for LLM Applications gives security guidance for LLM applications and includes some agent risks. The Agentic Top 10 adds guidance when the system operates as an agent. A system that uses Retrieval-Augmented Generation (RAG) is not always agentic: see When to use the Agentic Top 10.

OWASP Top 10 for LLM ApplicationsOWASP Top 10 for Agentic Applications
Scope"applications that use large language models"agents that "plan, decide, and act across multiple steps and systems"
EntriesLLM01:2025 to LLM10:2025 in the 2025 edition. LLM01:2026 to LLM10:2026 in the 2026 edition, with new numbers for most entries.ASI01 to ASI10
EditionVersion 2025 of November 18, 2024, which Modulos maps. The 2026 edition of August 4, 2026 is the "Current release". Version 2026, dated December 2025. OWASP made it available on December 9, 2025.
InformationOWASP Top 10 for LLM ApplicationsOWASP Top 10 for Agentic Applications

OWASP and AI governance frameworks ​

The OWASP LLM Top 10 2025 and Agentic Top 10 2026 do not map entries to ISO/IEC 42001 or EU AI Act articles. LLM04:2025 cites the NIST AI RMF as a related framework, without mappings to its Categories. The Agentic 2026 document has no mappings to those Categories. Appendix A of the LLM Top 10 2026 maps risks to NIST AI RMF Categories and other frameworks. See LLM mappings and Agentic mappings.

The GenAI Security Crosswalk is a different OWASP dataset that maps security risks to other frameworks. Its rule is: "Every mapping in this project is unreviewed until a named reviewer signs it."

Article 15 of the EU AI Act sets accuracy, robustness, and cybersecurity requirements for high-risk AI systems. Article 16(a) tells providers to "ensure that their high-risk AI systems are compliant with the requirements set out in Section 2". Article 15 applies from December 2, 2027 for Article 6(2) and Annex III systems, and August 2, 2028 for Article 6(1) and Annex I systems (Article 113(c)). For Article 6(1) systems related to Annex I Section B products, only Articles 6(1), 60a, and 102–112 apply, plus Articles 57–59 only insofar as the listed legislation incorporates high-risk requirements (Article 2(2)). See dates and transitional rules.

Other information from the OWASP GenAI Security Project ​

The OWASP GenAI Security Project publishes more than the two Top 10 documents. In the Letter from The Agentic Top 10 Leaders, the Agentic Security Initiative identifies these publications:

  • "Agentic AI - Threats and Mitigations", the "core taxonomy" of the Agentic Security Initiative
  • the "Agentic Threat Modelling Guide"
  • "Securing Agentic Applications"
  • "State of Agentic AI Security and Governance"

The OWASP GenAI Security Project also publishes the GenAI Security Crosswalk.

The OWASP AI Vulnerability Scoring System (AIVSS) is a different OWASP project. The OWASP Top 10 for Agentic Applications refers to AIVSS "for scoring and prioritization".

Frequently asked questions about OWASP for AI ​

For the frequently asked questions about the OWASP GenAI Security Project and these two Top 10 documents, see Frequently asked questions about OWASP for AI.

Sources ​