Center for Human-Compatible AI
The Center for Human-Compatible Artificial Intelligence (CHAI) is a multi-institution research group headquartered at UC Berkeley, founded in 2016 by Stuart Russell. CHAI's mission is to develop the conceptual and technical foundations needed to reorient AI research towards provably beneficial systems. The center's work spans value alignment, inverse reinforcement learning, human-robot cooperation, multi-agent systems, and models of bounded rationality. CHAI combines technical research with field-building through PhD training, internships, and annual workshops, and engages in policy advocacy on AI safety and autonomous weapons.
Funding Details
- Annual Budget
- -
- Monthly Burn Rate
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- $18,052,796
- Fiscal Sponsor
- University of California, Berkeley
Theory of Change
CHAI's theory of change centers on the idea that current AI systems are designed to optimize objectives specified by humans, but as these systems become more capable, even small misspecifications can lead to catastrophic outcomes. CHAI proposes a new model of AI in which machines are explicitly uncertain about human preferences, must learn those preferences from human behavior (via inverse reinforcement learning), and defer to humans rather than pursuing fixed objectives. By developing the mathematical and conceptual foundations for provably beneficial AI, training the next generation of AI safety researchers through PhD programs and internships, producing influential research, and shaping public policy through advocacy and public engagement, CHAI aims to redirect the trajectory of AI development toward systems that are fundamentally aligned with human values.
Grants Received
from Survival and Flourishing Fund
Projects
A research internship program within CHAI at UC Berkeley where participants work on AI safety research projects under the mentorship of PhD students and postdocs, typically resulting in first-author workshop papers.
People
No linked people.
Discussion
Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.
Details
- Last Updated
- Apr 2, 2026, 9:58 PM UTC
- Created
- Mar 18, 2026, 11:18 PM UTC