According to foreign media reports, this year's Fields Medal winner and mathematician Jacob Zimmerman recently announced the establishment of the Mathematics AI Safety Institute (MAISI), with Zimmerman serving as the institute's scientific director and Andrew Critch as secretary-general. The Fields Medal is awarded every four years to mathematicians under the age of 40 and is considered the most prestigious award in the field of mathematics.
MAISI to launch first semester of AI safety research in 2027
MAISI is modeled after the structure of the Institute for Advanced Study in Princeton and will invite mathematicians to conduct AI safety research on a semester basis. The institute plans to launch its first semester in January 2027, inviting 10 to 30 mathematicians to the Bay Area; in the special semester of September 2027, the scale will expand to 30 to 100 people. The goal of the institute is to develop concepts for definitions, measurement methods, and solutions aimed at determining the conditions under which one can be confident that specific AI systems will not cause disasters.
The advisory board includes Timothy Gowers, Ravi Vakil, Jeffrey Irving, and Paul Christiano. Among them, Gowers and Vakil are mathematicians, while Irving and Christiano have long been engaged in AI safety research. MAISI will collaborate with Alignment.org and Resolution.org.
Zimmerman will join the safety team at OpenAI this month, but MAISI stated that its operations are independent of OpenAI. Critch co-authored a paper with Zimmerman in 2025 that explored scenarios in which AI risks threaten all of humanity. The focus of AI safety discussions has previously been mainly on policy, evaluation, and experimentation, and it is not common for organizations to be established that bring together large numbers of mathematicians dedicated to theoretical foundations. Evan Hubinger, head of alignment science at Anthropic, said on the same day that a solution to the problem of super-intelligent alignment has not yet been found.