What is OWASP Top 10 for LLM Applications v1.0?

Apps built on large language models (the kind of AI behind chatbots) face a particular security problem. Because instructions and data reach the model together in the same text, the model cannot reliably tell them apart. Prompt injection is the attack that exploits this: a crafted input changes how the model behaves in ways its developer did not intend, and it can hide in a web page or document the model reads. OWASP, a nonprofit security foundation, publishes a Top 10 list of the most critical security risks for these applications, and its 2026 edition was published on August 4, 2026. The list is meant to help developers and security teams prioritize their security work.

Written from primary sources and reference pages, and checked against them by an independent AI fact-checker. Drafted with AI assistance.