Guide to Red Teaming Methodology on AI Safety
Japan AI Safety Institute guide to planning, conducting and reporting red-teaming of LLM systems
A methodology for red-teaming LLM-based systems: scoping, attack techniques (prompt injection, jailbreaks, data extraction), execution, evaluation and reporting, with roles and organisational preparation; one of the first government red-teaming guides.
Contribute
Spot an error, or know a record we should add? Members can propose additions and submit corrections.