Evan Hubinger is an AI safety researcher and Alignment Science lead at Anthropic. His work focuses on alignment research, stress-testing advanced AI systems, and understanding risks such as deceptive alignment and emergent misalignment. He previously conducted AI safety research at the Machine Intelligence Research Institute and OpenAI, and studied mathematics and computer science at Harvey Mudd College.

Unclaimed profile

Evan Hubinger is an AI safety researcher and Alignment Science lead at Anthropic. His work focuses on alignment research, stress-testing advanced AI systems, and understanding risks such as deceptive alignment and emergent misalignment. He previously conducted AI safety research at the Machine Intelligence Research Institute and OpenAI, and studied mathematics and computer science at Harvey Mudd College.

Career

Evan Hubinger studied at Harvey Mudd College (B.S. in Mathematics and Computer Science) and The College Preparatory School (High school diploma).

arxiv.org

Risks from Learned Optimization in Advanced Machine Learning Systems

Hubinger and collaborators introduce mesa-optimization and analyze why a learned optimizer’s goals may differ from its training objective.

Open ↗
arxiv.org

Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training

Hubinger and collaborators investigate whether deceptive behavior can survive standard safety-training techniques.

Open ↗

Evan's Collections

Evan's Lists

In these lists

Cosigns Given