New institute seeks mathematical proofs of AI safety
Canadian mathematician Jacob Tsimerman, a recent Fields Medal recipient, has announced the creation of the Mathematical A.I. Safety Institute, or MAISI. The independent research organization will be based in the San Francisco Bay Area and plans to begin work in January 2027 with a team of between ten and thirty mathematicians. Tsimerman is also joining OpenAI’s safety team.
Tsimerman says AI safety requires a far higher standard than is currently available. In cryptography, a scheme can be shown to resist every possible attack without testing each one. MAISI says AI currently offers no comparable shortcut: safety is assessed in practice, while the field still lacks a precise theoretical definition of what “safe” means.
The institute aims to develop methods for proving that AI systems behave responsibly and produce accurate results. It also wants to study whether groups of cooperating AI agents could create unintended outcomes and whether systems can withstand previously unknown vulnerabilities. One possible technique is zero-knowledge proofs, which could allow a system to demonstrate that it is not cheating without revealing an AI laboratory’s trade secrets.
Written by Sakhanda Wire AI from reporting by The Decoder. Facts, quotes and full context are in the original article below.