1. Home
2. News
3. OpenAI

The Association for Human Mathematics urges mathematicians to stop working with OpenAI

The group doesn't dispute a single proof. It objects to how OpenAI released them, while Vitalik Buterin worries about what the math means for encryption.

By Think Facility · October 8, 2026 · OpenAI Research Save

On October 6, 2026, OpenAI put hundreds of math manuscripts from an unreleased model on GitHub (our story on the release). The next day, the Association for Human Mathematics posted a short statement on its site asking mathematicians to stop working with the company.

The statement doesn't say any of the proofs are wrong. Its complaint is about how they arrived. Meanwhile a second argument started among people who guard cryptocurrency, about whether math this fast puts the keys to their wallets at risk.

Who

Association for Human Mathematics

About

OpenAI's October 6 Release of Mathematical Documents

Asks

mathematicians to discontinue their work with OpenAI

Buterin's view

lattices may take serious hits from the next two years of AI math

What the AHM objects to

The statement runs four paragraphs. It opens by noting that OpenAI is "currently defending lawsuits" over plagiarism, copyright and trademark claims, then says plainly that "Mathematicians did not ask for this work to be done."

Its main charge is about the advisory group OpenAI pointed to. According to the AHM, that group's first advisory statement said frontier AI companies shouldn't test advanced math problems on internal models at all, so OpenAI went ahead against the group's central premise.

The page doesn't list members or say how many mathematicians signed it. It ends by asking people to "return to a vision of science that centers human understanding."

Why cryptography came up

Scott Aaronson took a different line on his blog on October 7. He called it "one of the biggest days in mathematical history" (one of the claimed proofs is of the Unique Games Conjecture, which his wife has worked toward for as long as he's known her). Then he added an update pointing out that cryptography was missing from OpenAI's list.

He doesn't name the companies or the sources. Justin Drake had gone further that morning on X, asking the blockchain industry to "calmly begin planning" for what he calls bunker mode: moving funds to fresh addresses whose public keys have never been exposed. He wrote that ECDSA could break before quantum computers do, "in the worst case in months not years."

Buterin's answer

Vitalik Buterin replied at 7:28 pm Eastern. He agreed the risk is worth taking seriously but told people not to panic, and he widened it past elliptic curves to lattices, which most people had treated as safe.

His advice is mostly about size. For anything lattice-based, he says to be much more paranoid about parameters, and for long-term secrets he suggests multiplying key sizes by 10. Hash-based schemes, which are designed to have no structure to exploit, are his preference wherever they work.

He's also wary of the fix Drake proposed. Fresh addresses are a good idea if they're easy for you, Buterin wrote, but he has "lost more money in botched migrations" than in every hack combined.

OpenAI hasn't answered the AHM on any page we saved. Drake says he'll take questions from institutions in London next month.