OWASP/www-project-top-10-for-large-language-model-applications

Provides a curated risk framework specifically tailored to LLM applications, analyzing how traditional vulnerabilities manifest differently in generative AI contexts rather than simply adapting existing OWASP guidance. The project bridges general application security principles with LLM-specific attack surfaces, offering developers practical remediation strategies adapted for prompt injection, model poisoning, and other AI-native threats. Built as a living standard under the broader OWASP GenAI Security Project with community-driven updates and a v2.0 roadmap.

1,145 stars.