Skip to main content
    Skip to main contentSkip to navigationSkip to footer
    Artificial Intelligence

    OWASP LLM Top 10

    Also known as:
    LLM Top 10
    OWASP AI Security
    LLM Risk List
    OWASP LLM Risks
    Updated: 2/9/2026

    A standardized list of the most critical security risks for LLM applications, published by OWASP.

    Quick Summary

    OWASP LLM Top 10 is the industry standard for LLM security risks: Prompt Injection (#1), Insecure Output (#2), Data Poisoning (#3). Required reading for AI developers.

    Explanation

    The Top 10: 1) Prompt Injection, 2) Insecure Output Handling, 3) Training Data Poisoning, 4) Model DoS, 5) Supply Chain Vulnerabilities, 6) Sensitive Info Disclosure, 7) Insecure Plugin Design, 8) Excessive Agency, 9) Overreliance, 10) Model Theft.

    Marketing Relevance

    OWASP LLM Top 10 is the reference for LLM security audits. Helps with risk assessment, compliance documentation, developer training.

    Example

    A security review uses OWASP LLM Top 10 as a checklist: Is prompt injection mitigated? Are outputs sanitized? Is training data secure?

    Common Pitfalls

    List is a snapshot – new risks emerge. Not all risks equally relevant for every use case. Detail level for implementation often insufficient.

    Origin & History

    OWASP LLM Top 10 v1.0 was released August 2023, v1.1 in October 2023. The project has 500+ contributors. Version 2.0 is in progress for 2025.

    Comparisons & Differences

    OWASP LLM Top 10 vs. OWASP Top 10 (Web)

    Classic OWASP Top 10 covers web app risks (SQL injection etc.); LLM Top 10 covers AI-specific risks.

    OWASP LLM Top 10 vs. MITRE ATLAS

    OWASP LLM Top 10 is prioritized risk list; MITRE ATLAS is comprehensive framework with techniques and tactics.

    Related Services

    Related Terms

    LLM SecurityPrompt InjectionAI Safetysecurity-auditrisk-assessment
    👋Questions? Chat with us!