Nonprofit Trillium Labs Launched for AI Research

The new research organization aims to advance recursive self-improvement and agentic AI through open transparency.

Updated on Oct. 2, 2026 in Artificial Intelligence

Bold flat-color editorial illustration of a glass prism anchored in metallic cubes, evoking the structure of transparent artificial intelligence research. — Nonprofit Trillium Labs has launched with backing from Schmidt Sciences and Halcyon Futures to advance transparent, open-source research on agentic AI models. AI Illustration. Upload story photo >

Live Poll

Should artificial intelligence labs be required to make their experimental research public?

Nathan Lambert and Tom Zick have established Trillium Labs, a nonprofit dedicated to researching recursive self-improvement and post-training methods for artificial intelligence. By publishing experimental details, the group seeks to bridge the resource gap between industrial development and academic study.

Why it matters

The founders believe that transparent, open-source research is essential for effective community scrutiny and the mitigation of risks associated with advanced AI models. This approach aims to address the growing disconnect between proprietary industry advancements and independent academic inquiry.

The organization plans to dedicate $30 million toward model training over the next 18 months. Research efforts will utilize reinforcement learning techniques to refine both behavioral outcomes and technical model capabilities.

The players

Nathan Lambert

He is a co-founder of Trillium Labs and a researcher with a background in advanced artificial intelligence studies.

Tom Zick

He is a co-founder of Trillium Labs who focuses on recursive self-improvement and post-training research methods.

Schmidt Sciences

This philanthropic organization provides funding and support for scientific and technological research initiatives.

Halcyon Futures

This investment group provides capital to organizations focused on long-term technological and societal development.

The details

Trillium Labs will operate with financial backing from Schmidt Sciences and Halcyon Futures to support its experimental agenda. The nonprofit intends to share its findings publicly to ensure outside scientists can replicate and verify its research on agentic AI.

Trillium Labs officially launched on October 2, 2026.
The organization plans to spend $30 million on training over the next 18 months.

The Tech Race

Trillium Labs enters an industry dominated by proprietary research, effectively challenging the closed-source norms of major AI laboratories. By prioritizing open academic standards, the nonprofit aims to steer the technical trajectory of agentic AI toward greater safety and transparency.

The organization's commitment to open research means that developers and academics will gain greater access to standardized, replicable data on AI behavior. This transparency could lead to safer, more reliable AI tools that eventually reach the consumer market.

The takeaway

Trillium Labs represents a growing movement to reclaim AI research for the academic and public community. Users should monitor these open research outputs as indicators of future safety standards for AI agents.