Sandhini Agarwal is an AI policy and safety researcher whose work has included bias analysis, content filtering, deployment evaluations, red teaming, and governance of agentic systems. At OpenAI, she led launch safety for GPT-4 and chaired the Safety Advisory Group in 2025. She holds a B.A.S. in Philosophy and Symbolic Systems from Stanford University.
Started on the Deployment Planning team: led GPT-3 bias analysis and built the first content filter. Later led GPT-4 launch safety, worked on deployment evals, red teaming and agentic-systems governance, and chaired the Safety Advisory Group in 2025.
TPM Intern
B.A.S., Philosophy and Symbolic Systems
Undergraduate studies in Symbolic Systems and Philosophy
