What Center for Human-Compatible AI does

The Center for Human-Compatible AI (CHAI) is a multi-institution AI safety research group based at UC Berkeley, founded in 2016 with the stated mission of developing the conceptual and technical wherewithal to reorient AI research toward provably beneficial systems. The center's core research spans five areas: value alignment and inverse reinforcement learning (teaching AI systems to infer human preferences from behavior), human-robot cooperation, multi-agent systems cooperation, foundations of rational agency and causality, and models of bounded rationality. Notable technical contributions include the Cooperative Inverse Reinforcement Learning (CIRL) framework (NeurIPS 2016), the Off-Switch Game for corrigibility, and more recent work on political neutrality evaluation (Maximum Equal Approval algorithm) and the Learning to Yield and Request Control (YRC) coordination benchmark.

CHAI operates as a non-commercial academic research center rather than a product company. Its funding comes from academic sponsors (Open Philanthropy, Future of Life Institute, Leverhulme Trust, CITRIS) and public donations, not from selling products or services. The center's faculty includes Stuart Russell (author of the leading AI textbook), Pieter Abbeel, Anca Dragan, alongside Senior Scientist Jonathan Stray and Executive Director Mark Nitzberg, with affiliate faculty at Oxford, Cambridge, Cornell, and Michigan. Research outputs are disseminated through open-access publications (arXiv, top-tier ML conferences), GitHub repositories, an email newsletter, and an annual workshop series.

CHAI's audience comprises the academic AI research community (its primary segment), governments and international organizations seeking AI policy guidance, and AI labs and technology organizations seeking alignment advisory input. The center also runs a structured research internship program (in-person at $5,000/month, remote at $3,500/month) that alumni channel into influential AI safety research positions, and it advises governments on topics including lethal autonomous weapons and international AI governance.

Name

Center for Human-Compatible AI

Legal name

UC Berkeley Center for Human-Compatible AI

Website

https://humancompatible.ai

Company type

Private

Founded year

2016

Operating status

Operating

Headcount range

11–50 employees

Short description

The Center for Human-Compatible AI (CHAI) is a UC Berkeley-affiliated multi-institution research center founded in 2016 that conducts foundational research on AI alignment, value learning, corrigibility, and human-robot cooperation, serving the academic AI safety community, policymakers, and AI labs.

Ownership category

akta.pro rank

Headquarters

HQ city

Berkeley

HQ country

United States

HQ region

North America

Offices1 record

Markets served

GTM type

B2B

Offering type

Services

Cost components

Personnel, Technology or R&D, Operations, Marketing or Sales

Revenue model

1. Research Grants and Sponsorship: The center is funded by academic sponsors including Open Philanthropy, Future of Life Institute, Leverhulme Trust, and CITRIS (Center for Information Technology Research In The Interest of Society). These provide operational funding for research activities, faculty, and fellows.
2. Donations: The center accepts donations through its website to support ongoing research efforts in beneficial AI development.

Pricing tiers

Go-to-market motion1 record

Distribution channels2 records

Marketing channels5 records

Core offering

The Center for Human-Compatible AI (CHAI) is a multi-institution academic research group based at UC Berkeley that conducts foundational research on AI safety and alignment. Its work spans value alignment, inverse reinforcement learning, human-robot cooperation, multi-agent systems, rational agency, and bounded rationality, producing peer-reviewed publications, open-source benchmarks, and policy advisory outputs rather than commercial products.

Product overview

The Center for Human-Compatible AI (CHAI) is a research organization at UC Berkeley focused on developing beneficial AI systems. Rather than offering commercial products, CHAI produces academic research publications, maintains a comprehensive annotated bibliography of AI safety materials, and offers research training programs including internships and fellowships. Their core offering is foundational research into AI alignment, value learning, human-robot cooperation, and safety verification.

Differentiator

Problem solved

Functional benefit

Products and services

• Research Internship Program Paid research internship program offering in-person roles at $5,000/month and remote roles at $3,500/month, typically 12 to 16 weeks, designed to train future AI safety researchers. Open to international applicants.

Quantifiable outcome

• Hundreds of peer-reviewed publications across AI alignment, value learning, human-robot interaction, and related fields since 2016
• +1 more outcomes

Segments3 records

Ideal customer profiles3 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability10 records

Feature5 records

Partnerships

Eight partnerships are on record, tiered core and major.

• CITRIS (Center for Information Technology Research In The Interest of Society)coreStrategic or Co-development PartnerUC Berkeley research center that sponsors CHAI, facilitating collaboration on societal-scale technology challenges including AI safety.
• Leverhulme Centre for the Future of IntelligencemajorStrategic or Co-development PartnerPartner organization at Cambridge University focused on the future of AI and its implications for humanity, collaborating on shared research interests in AI safety and governance.
• Center for Long-Term CybersecuritymajorStrategic or Co-development PartnerUC Berkeley center focused on long-term cybersecurity implications, partnering with CHAI on research at the intersection of AI and security.
• Berkeley Existential Risk InitiativemajorStrategic or Co-development PartnerBerkeley-based initiative focused on existential risk from advanced AI systems, collaborating with CHAI on safety research.
• Berkeley Kavli Center for Ethics, Science, and the PublicmajorStrategic or Co-development PartnerPartnership focused on ethical dimensions of AI development and public engagement with AI governance issues.
• ICT4Peace FoundationmajorStrategic or Co-development PartnerInternational foundation focused on the role of ICT in peace and security, partnering with CHAI on governance and policy dimensions of AI safety.
• Centre for the Study of Existential RiskmajorStrategic or Co-development PartnerCambridge University center studying existential risks including advanced AI, collaborating with CHAI on shared research priorities.
• Future of Humanity Institute (Oxford)majorStrategic or Co-development PartnerOxford University's research institute on existential risk, partnering with CHAI on technical AI safety research and policy analysis.

Scale indicators2 records

Recent moves6 records

Expansion highlights5 records

Direct peers

• Machine Intelligence Research Institute (MIRI): An AI safety research organization focused on alignment and existential risk from advanced AI. Highly comparable to CHAI as a peer nonprofit AI alignment research entity, with overlapping intellectual scope in formal alignment and value learning research.
• Redwood Research: An AI safety research nonprofit working on alignment and deceptive behavior in frontier models. Directly comparable to CHAI as a focused AI alignment research organization publishing papers and advising on practical alignment problems.
• Centre for the Study of Existential Risk (CSER), Cambridge: Cambridge University research center studying existential risks including advanced AI. Directly comparable as a multi-disciplinary academic research center with overlapping risk/AI safety mandate; listed as a CHAI partner.

Broad incumbents

• Anthropic: A frontier AI lab structured around long-term AI safety, with significant in-house alignment, interpretability, and constitutional AI research. Comparable to CHAI in research priorities around alignment, though operating as a large commercial lab with the talent concentration (including Anca Dragan) that CHAI depends on academically.
• OpenAI (with focus on alignment/safety team): Frontier AI lab with a dedicated alignment and safety research organization. Comparable in addressing the same technical problems (alignment, value learning, oversight) that CHAI researches, but at substantially larger scale and with commercial deployment.
• Google DeepMind (Safety & Alignment teams): Major AI lab with dedicated safety and alignment research groups (e.g., on RLHF, mechanistic interpretability, scalable oversight). Operates in the same intellectual space as CHAI's alignment work but at industry scale and compute.
• Future of Life Institute (FLI): International nonprofit funding AI safety research and policy advocacy. Comparable as both an AI safety ecosystem counterparty to CHAI and as an overlapping funder/advocate; explicitly a CHAI sponsor.
• UK AI Safety Institute: Government-backed institute conducting pre- and post-deployment evaluations of frontier AI. Comparable to CHAI in mandate (independent AI safety evaluation and governance advice) and increasingly a counterpart body on policy advisory work CHAI conducts.

Regional players

• Berkeley AI Research (BAIR): UC Berkeley's broader AI research lab that CHAI is structurally integrated with. Highly comparable as overlapping research community at the same institution, sharing faculty and intellectual milieu.

Emerging players

• Conjecture: An AI safety startup focused on aligning and containing frontier AI systems. Comparable as a smaller, mission-driven AI safety organization pursuing alignment research, though structured commercially rather than academically.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat4 records

Key risks6 records

Key highlights7 records

Customer concentration

Revenue estimate

Valuation estimate

Number of profiles

Profiles5 records

Funding overview

Funding rounds

Investors

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Center for Human-Compatible AI