OpenAI's Approach to External Red Teaming for AI Models and Systems

Lama Ahmad, Sandhini Agarwal, Michael Lampe +1

Red teaming has emerged as a critical practice in assessing the possible risks of AI models and systems. It aids in the discovery of novel risks, stress testing possible gaps in ex…