Proof verification took a notable step on 26 October 2026. Axiom, a San Francisco AI company, says its AI agents produced the first complete computer check of the Classification of Finite Simple Groups. The company published the claim in a blog post.
What Axiom says happened
The classification is one of the longest proofs in mathematics. Axiom says the original argument runs to about 15,000 pages across more than 500 articles by about 100 mathematicians.
Axiom says it rewrote the argument in Lean, a language that software can check line by line. Swarms of AI agents did the work in 24 days. The post puts the result at about 43.1 million lines of Lean. The project was split into 108 milestones and about 380 smaller tasks.
Axiom says it formalized the first generation proof. The second generation proof is still incomplete.
Errors found along the way
Axiom reports 34 noteworthy deviations from the published record. It also lists minor typos and table errors. Axiom says the agents found a counterexample to a lemma in the quasithin volumes, and counterexamples to a closure assertion in a 1986 treatment of the Unbalanced Group Theorem.
Axiom says it closed that gap with a new result, and that group theorists Richard Lyons and Ronald Solomon were consulted. The post also says some agent results rested on unproved assumptions. Lean accepted them, but they left the goal unfinished.
What it means for people who check data
The lesson is not about group theory. It is about where trust sits in a checked result.
Axiom says Lean checks each deduction. Humans must still confirm that the formal statement says what the classification means. A passing check proves the statement that was written. It does not prove that the right statement was written. Axiom says human mathematicians judged the consequential errors and repairs.
The same split applies to business data. A rule can confirm that a number follows from its inputs. A person still has to confirm the inputs and the question were right. Our guide to data verification covers that step, and data validation explains how a format check differs from a truth check. The trust index shows how we score evidence.
What to keep in mind
All figures here come from Axiom's own post. Press coverage repeats them. We found no independent review yet. Axiom frames the result as the longest proof in the history of mathematics, and that framing is its own. Wait for outside mathematicians to review the formal statement before treating the claim as settled.