OpenAI Faces Mathematician Backlash Over AI Math Claims

OpenAI, Anthropic and other artificial intelligence labs have announced breakthroughs on numerous long-standing mathematical problems over the past year, in some cases surpassing what researchers expected current systems could do, including resolving one of the Millennium Prize problems, according to an Oct. 5 report from The Verge. The results, which might once have drawn unqualified praise, have instead triggered sustained criticism from mathematicians over how the companies rolled them out, the outlet reported.

By the numbers, per The Verge's reporting:

• 1 — Millennium Prize problem an AI system has reportedly resolved.
• 3 — Dated dispatches, published Sept. 23, Sept. 28 and Oct. 5, that The Verge has run tracking the dispute so far.

What breakthroughs set off the dispute?

The Verge is tracking the story as a running package of dispatches, with installments titled "OpenAI keeps bulldozing mathematicians," "OpenAI just wants to win" and "The AI takeover of mathematics has begun." Those headlines capture a pattern the outlet describes: OpenAI producing genuinely notable mathematical results, then mismanaging how it tells the world about them. Reporter Robert Hart wrote that the company has shown it can produce impressive breakthroughs and then badly fumble their announcement, a cycle the outlet says repeated itself over several months.

That repetition, according to The Verge's Sept. 23 dispatch, turned a string of mathematical results into what the outlet called a "reputational crisis" for OpenAI. The company responded by beginning to consult outside mathematicians about how to handle future disclosures, The Verge reported.

What is OpenAI's new advisory group supposed to do?

OpenAI's latest fix, detailed in a Sept. 28 Verge dispatch, is a new independent advisory group made up of elite mathematical practitioners. Its mandate, per that reporting, is to help the company coordinate the release of further results — a task complicated by OpenAI's own disclosure that an unreleased model has already generated a large batch of additional findings awaiting publication. The Verge reported that prospect is already stirring dread among researchers over what a continued wave of machine-produced mathematics could do to the field.

Why do mathematicians say the process is still broken?

Mathematicians who spoke with The Verge, including one member of the new advisory group, described the rollout as "messy and confusing," the outlet reported. Their account, relayed by The Verge, said the process carries many of the same hallmarks as OpenAI's earlier rushed forays into mathematics, suggesting the company has made little progress since its earlier missteps.

OpenAI, for its part, says it is trying to do better. The Verge's framing of that claim is skeptical, noting the company's record over the preceding months and leaving open whether the promise will hold up.

How does this fit into the wider AI push into mathematics?

OpenAI is not working this frontier alone. The Verge's reporting notes that Anthropic and other labs have also claimed advances on problems mathematicians once expected to resist automated solutions for years, if not longer. The reported resolution of one Millennium Prize problem stands as the most prominent example cited in that coverage, underscoring how quickly expectations for AI's mathematical capability have shifted.

The dynamic described in The Verge's coverage is one of Silicon Valley's broader move-fast instincts colliding with a discipline built on slow, peer-reviewed verification — producing results researchers might otherwise celebrate without reservation, if not for the way they have been announced.

What happens next for OpenAI's math results?

The next test for OpenAI and its advisory group is how they handle the release of the additional results now awaiting disclosure. Mathematicians are watching that process warily, according to The Verge, given the history recounted across its coverage so far.

For now, two competing narratives run through the reporting: a research frontier advancing faster than many mathematicians anticipated, and a disclosure process that keeps drawing criticism from the people best positioned to judge the underlying work. The Verge's running coverage suggests both will remain true for some time.