← AI Glossary

OWASP Top 10 for LLM Applications

Sovereignty & security

The OWASP Top 10 for LLM Applications is the reference ranking of the ten most critical security risks for applications built on a large language model. It is published by the GenAI Security Project of OWASP, a non-profit foundation dedicated to software security. Its 2025 edition, released on 17 November 2024, lists in order: 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 2026 edition, published in August 2026 and for the first time cross-checking the community vote against an analysis of 6,639 real incidents, keeps prompt injection at the top, moves excessive agency up to third place and broadens system prompt leakage into “hidden context exposure”. For agents that act on tools, OWASP has also published a separate Top 10 for agentic applications since 9 December 2025.

In practice at Gensai

As part of enhanced security testing, Gensai uses the OWASP Top 10 to review a project's risks from the scoping stage, then to build part of its test scenarios.

What is this ranking for, in a business?

As a checklist, at the scoping stage and then during testing: for each risk, check whether it applies to the use case and which measure covers it. It also provides a common language with the IT department, the CISO or an auditor. For a chatbot that answers from documents, prompt injection, sensitive information disclosure and embedding weaknesses are often the first points to examine. For an agent that acts on tools, the Top 10 for agentic applications should be added.

Common mistake

Treating the Top 10 as a standard or a certification. It is a ranking of the most frequent and most serious risks, revised from one edition to the next: an application can cover all ten points and still be vulnerable to a risk specific to its context.

Definition written with AI assistance, checked by Sébastien Lecocq, co-founder of Gensai.

Going beyond the definition?

From concept to project: Gensai builds custom, sovereign, GDPR-compliant AI solutions.

Related services: Chatbots and answer engines, Custom AI solutions

55 boulevard de Strasbourg, 75010 Paris
Information used only to answer the request.