Why it matters
Since 2017, BERI has been the operational backbone for existential risk research that wouldn't happen otherwise. Through fiscal sponsorship and university collaborations, we've enabled research groups, independent scholars, and project teams to receive funding, hire staff, and operate with the agility their work demands.
We don't run these programs — we make it possible for the experts to run them well.
Testimonials
66
In 2023, the Center for Human-Compatible AI (CHAI) at UC Berkeley faced a familiar research bottleneck: They needed $800,000 in specialized computing hardware to train LLMs and advance alignment research, but UC Berkeley's procurement system moves at university speed, not AI development speed.
University procurement requires extensive documentation, multiple approval layers, and adherence to contracting rules designed for traditional academic timelines. For cutting-edge AI safety research, this mismatch could mean months of delay while the field moves forward.
BERI bypassed this entirely. Using our flexible nonprofit structure, we purchased and delivered the computing infrastructure directly. What would have taken UC Berkeley's system months took BERI weeks, allowing CHAI researchers to move forward with critical research without losing momentum.
The Broader Picture
CHAI's research directly informs how major AI labs approach alignment problems. Every month of delay in their work is a month where those insights aren't available to the broader field. BERI's ability to move at the speed of AI development, rather than institutional bureaucracy, means safety research keeps pace with capability development.
This is a BERI collaboration working as it should: Researchers focus on the science while BERI handles the operational friction that would otherwise slow them down.