The Mathematical AI Safety Institute wants to prove AI is safe the way cryptographers prove codes are unbreakable
decoder ·
Jacob Tsimerman, a Canadian mathematician and recent Fields Medal recipient, announced the founding of the Mathematical A.I. Safety Institute (MAISI). MAISI is an independent research institute in the San Francisco Bay Area that plans to begin work in January 2027 with ten to thirty mathematicians addressing AI safety problems, according to the New York Times. Tsimerman, who is also joining OpenAI's safety team, said the field needs a much higher safety standard than it currently has. The article contrasts cryptography, where an encryption scheme can be proved unbreakable without trying every possible attack, with AI, which lacks such a shortcut. According to MAISI, safety only shows up in practice, and there is not even a clear theoretical definition of what safety means. MAISI wants to make it possible to prove that a system acts responsibly and produces correct results, that multiple AI agents working together do not trigger unwanted outcomes, and that systems can withstand vulnerabilities nobody has found yet. One possible tool is zero-knowledge proofs, which could let a system demonstrate that it is not cheating without exposing AI labs' trade secrets.
decoder