Sandhini Agarwal researches AI safety and governance after leading launch-safety work for GPT-4 at OpenAI. She co-authored OpenAI's practices for governing agentic systems and contributed to model safety evaluations, red teaming and misuse research.
Sandhini Agarwal researches AI safety and governance after leading launch-safety work for GPT-4 at OpenAI. She co-authored OpenAI's practices for governing agentic systems and contributed to model safety evaluations, red teaming and misuse research.
Career
AI safety and policy researcher at OpenAI until 2026-07Worked on AI safety, deployment evaluations and governance; led launch safety for GPT-4.