Jacob Tsimerman, the 2026 Fields Medal winner, has founded the Mathematical AI Safety Institute, a new research body dedicated to building rigorous mathematical frameworks for understanding and mitigating the risks of artificial intelligence. The institute, known as MAISI, was launched on September 8, 2026.
Tsimerman will serve as MAISI’s Scientific Director while simultaneously joining OpenAI’s safety department in September 2026. The institute, however, will operate independently from the company.
What MAISI actually plans to do
Based in the Bay Area, MAISI is modeled after traditional mathematical research institutes where small groups of serious researchers gather for structured semesters to work on hard problems.
The institute plans to begin operations in January 2027 with an initial cohort of 10 to 30 mathematicians. A more ambitious “special year” program is slated for September 2027, scaling up to 30 to 100 researchers for an intensive research period.
Andrew Critch, who co-authored a 2025 paper with Tsimerman on what they termed “omnicidal” scenarios related to AI dangers, will serve as Executive Director.
The advisory panel includes Stanford’s Ravi Vakil, and MAISI has partnered with organizations like alignment.org and resolution.org. Applications for mathematicians who want to join are already live on the institute’s website.
Why a mathematician is leading AI safety research
Tsimerman’s Fields Medal, awarded in July 2026, recognized his groundbreaking work on problems including the André-Oort conjecture. He was the first faculty member at a Canadian university to receive the honor. The medal is awarded only once every four years to researchers under 40.
Tsimerman’s 2025 paper on omnicidal AI scenarios with Critch appears to have been the intellectual catalyst for MAISI. The paper laid out formal frameworks for understanding how AI systems might cause catastrophic harm, the kind of structured analysis that a full institute could build upon.
The partnership with alignment.org and resolution.org suggests MAISI won’t operate in isolation.
The institute’s phased scaling plan, from a small initial group to potentially 100 researchers within its first year of expanded operations, would make it one of the largest dedicated AI safety research bodies in the world.