Schmidt Sciences Names 28 AI2050 Fellows, Backed by Up to $16 Million

Who was selected and how much funding is involved

The nonprofit Schmidt Sciences announced the 2026 AI2050 cohort on October 1, 2026. The 28 researchers come from around the world and will be supported by up to $16 million over three years to pursue their work. According to the organization, the addition brings the AI2050 community to 127 fellows representing 50 institutions across 11 countries, with up to $80 million in funding committed to date.

The six senior fellows are Francis Bach (Inria), Mikhail Belkin (University of California San Diego), David Danks (University of Virginia), Sendhil Mullainathan (Massachusetts Institute of Technology), Alondra Nelson (Institute for Advanced Study) and Bernhard Schölkopf (ELLIS Institute Tübingen and Max Planck Institute for Intelligent Systems).

The 22 early-career fellows include researchers at Carnegie Mellon University, MIT, Cornell, Princeton, ETH Zurich, the University of Toronto, Nanyang Technological University and Bayero University, Kano, among others. Trillium Labs and Aithrya are also represented.

What the fellows will work on

Schmidt Sciences says the awardees will benchmark the capabilities and risks of self-improving AI systems, use AI to model and predict extreme weather events that have never been recorded but remain physically possible, and design AI systems to guide experiments aimed at engineering immune cells to fight cancer and autoimmune diseases.

The senior fellows' projects cover a broad range:

• Francis Bach will develop an external auditing system that checks whether AI agents' actions follow safety rules and identifies when they can be trusted to act reliably.
• Mikhail Belkin will investigate the hidden internal mechanisms that drive complex AI behavior.
• David Danks will study how human values and AI systems change one another over time and turn the findings into practical guidance for design, development and governance.
• Sendhil Mullainathan will build tools that translate patterns detected by AI into clear, testable scientific theories across medicine, chemistry and education.
• Alondra Nelson will follow three American families across three regions and three generations to document how overlapping AI systems shape everyday life.
• Bernhard Schölkopf will develop models that let AI systems learn by observing, acting and experimenting, with the aim of improving reasoning about interventions and robust generalization.

UC Berkeley said its assistant professor Irene Chen will focus on studying the hidden failures of AI when used for important decisions. Berkeley also noted that six of the fellows are EECS alumni: Bach, Andrea Bajcsy, Daniel Fried, Nathan Lambert, Juan Carlos Perdomo and Romain Lopez.

About the AI2050 program

Launched in 2022, AI2050 starts from a question: "It's 2050. AI has turned out to be hugely beneficial to society. What happened?" Grantees focus on a working list of Hard Problems in AI, from making sure the technology does not endanger humans to finding ways it can address global issues. The program was conceived and is co-chaired by Eric Schmidt and James Manyika, SVP of Research, Labs, Technology & Society at Google.

Beyond money, fellows attend an annual gathering to share findings and build connections across disciplines. They may also gain access to additional funding for collaborations with other fellows, as well as funding for substantial compute needs.

Over its first four years, the program has supported fellows who developed AI tools to help power grids handle the changing supply of wind and solar power, built an AI system to design novel molecules that chemists can make in the lab, and showed how AI-powered computer worms could autonomously adapt their attacks to each new target, giving cybersecurity defenders an early warning.

What the leaders say

"AI will shape nearly every aspect of our future, but the direction it takes is not inevitable," said Eric Schmidt, co-founder of Schmidt Sciences and co-chair of AI2050, according to the announcement. James Manyika argued that the key questions about AI "cannot be answered by computer science alone" and require insights from the sciences to the humanities.

The cohort's mix of safety, interpretability, scientific discovery and social research reflects that interdisciplinary framing. For the broader field, the announcement signals continued philanthropic funding for work on AI risks and benefits, with cumulative commitments now reaching up to $80 million.