OWASP Top 10 for Large Language Model Applications

The OWASP Top 10 for Large Language Model (LLM) Applications is a written guidance document developed by the OWASP community to identify the most critical security risks affecting applications that use large language models. 

First introduced in 2023 and updated in 2025, it highlights ten key vulnerability categories specific to AI systems, such as prompt injection, sensitive information disclosure, and data poisoning. The document is intended for developers, security professionals, and organizations building or deploying LLM-based applications.

For each risk, it provides a description of the vulnerability, examples of possible attack scenarios, and recommended prevention and mitigation strategies. The list is created through a community-driven process involving developers, data scientists, and security experts from around the world. Its goal is to raise awareness about emerging AI security threats and provide practical guidance for building safer AI systems. The OWASP Top 10 for LLM Applications does not impose rules or certification requirements but serves as an informational resource and best-practice reference. By identifying the most important risks, it helps organizations prioritize security efforts when designing and deploying LLM applications. Overall, the document contributes to promoting more secure and responsible use of large language models.