Quick answer
The OWASP Top 10 for LLM Applications is a widely used list of the most critical security risks in applications built on large language models, published by the OWASP GenAI Security Project. The 2025 edition covers prompt injection, sensitive information disclosure, supply chain, data and model poisoning, improper output handling, excessive agency, system prompt leakage, vector and embedding weaknesses, misinformation and unbounded consumption.
The Ten Risks
- Prompt injection: inputs that override the system's instructions. See prompt injection.
- Sensitive information disclosure: the model revealing personal, confidential or proprietary data.
- Supply chain: risks from third-party models, datasets, plugins and libraries.
- Data and model poisoning: manipulated training or retrieval data. See RAG and data poisoning.
- Improper output handling: passing model output to other systems without validation, enabling classic attacks such as cross-site scripting.
- Excessive agency: AI with more permissions or autonomy than it needs. See excessive agency.
- System prompt leakage: exposing instructions that contain business logic or secrets.
- Vector and embedding weaknesses: flaws in how RAG systems store and retrieve data, including permission bypass.
- Misinformation: confident but wrong output that users or systems rely on.
- Unbounded consumption: abuse that drives up cost or exhausts resources.
How to Use It
Treat the list as a testing checklist for any AI feature before release, and as a common language between developers, security teams and auditors. It complements, rather than replaces, standard web application testing against the OWASP Top 10, because AI features are still web applications underneath.
Testing Against It
An LLM application penetration test covers each category alongside traditional web and API testing, and maps every finding to the relevant entry.
Regulatory and Framework Context
For US organizations, mapping AI findings to the OWASP list makes them easy to align with the NIST AI RMF and ISO/IEC 42001.
Latest intel on this topic
- GenAI and Agentic AI Exploit Roundup Q3 2026OWASP GenAI Security Project · October 8, 2026
- HITCON CMT 2023 - LLM Security Presentation and Trip ReportEmbrace The Red · September 18, 2023
Updated hourly from the daily briefing.
Common questions
Is the OWASP Top 10 for LLM Applications a standard?
It is community guidance rather than a formal standard, but it is widely used by developers, testers and auditors as the reference list of AI application risks.
How often is it updated?
OWASP has revised it as AI applications evolved, with a major 2025 edition. Check the OWASP GenAI Security Project for the current version.
Change log
- November 13, 2026First published.