Which mentor is the AI Safety mentor known for scalable oversight in AI safety?

• A. Sam Bowman
• B. Peter McCrory
• C. Saffron Huang
• D. Nicholas Carlini

Answer: (A)

Scalable oversight in AI safety is about supervising increasingly capable models in a way that keeps pace with their abilities. It focuses on creating scalable methods for evaluation and guidance, such as using human feedback signals, preference learning, reward modeling, and task decomposition so supervision remains reliable even when a single human can’t assess every possible outcome.

Sam Bowman is identified here as the mentor known for scalable oversight because his work centers on building evaluation and supervision frameworks that scale with model performance, helping ensure models stay aligned through scalable feedback and measurement. This focus distinguishes him from the other listed mentors, who are associated with different aspects of AI safety or machine learning (such as adversarial robustness or broader ML research).