AI Companies Come for Mathematics
September 14, 2026
This week, OpenAI claimed that a swarm of its agents—about ten times as many as involved in July’s Hugging Face security breach—solved one of the million-dollar Millennium Prize problems, some of the most famous and difficult open problems in mathematics. The announcement generated enormous controversy after two mathematicians working on a closely related problem suggested OpenAI may have effectively cheated, relying on their unpublished research. In response, many of the world’s most prominent mathematicians—including some who were previously much more optimistic about the impact of AI on the field—published a statement decrying these companies’ “severe” and “general threat to intellectual work.”
In a prescient essay from our Summer 2026 issue that has gone viral, Columbia University mathematician Michael Harris, author of Mathematics Without Apologies: Portrait of a Problematic Vocation, explains it all: the long history and current stakes of AI’s designs on mathematics, why the field “only makes sense as a gift economy,” how “trade secrets” compromise human understanding, and what he hoped to achieve in helping to draft June’s Leiden Declaration, which warned about the issues at the center of this week’s controversy and called on the mathematical community to take action.
Here’s an excerpt from Harris’s piece:
They are interested in problems like the Riemann Hypothesis as a way to get your attention—your attention, dear reader!—because, although the million dollars of the Millennium Prize may be less than what Google spent on AI every ten minutes in 2025, it is still the kind of trophy that can consecrate its winner with cultural capital—which, by a miraculous transmutation, translates into colossal investments of actual capital. . . .>
The Angel of History looks on, puzzled, as the press releases pile up. An industry that sees thinking machines as debugged versions of human beings makes increasingly bold claims for their models’ performance on a series of increasingly ambitious benchmarks. First Proof’s authors have criticized submissions from OpenAI and DeepMind for their lack of transparency, especially the latter’s “undisclosed human guidance.” Even Aletheia’s March proof, presented as completely autonomous, was the result of a training process whose details are veiled by commercial secrecy and nondisclosure agreements. How much of its claimed autonomy derives from the scaffolding produced by human mathematicians, these “people-as-a-service”?
“Substituting a monetizable ersatz culture for the one evolved through historical experience,” Harris concludes, “is a particularly insidious form of knowledge collapse.” We believe in gift economies too, which is why you can read Harris’s whole essay—which readers have called “extraordinary,” “beautiful,” “the best essay on AI and knowledge I’ve read,” “the most cogent public statement by a prominent mathematician”—for free on our website. But as a small, independent nonprofit, we rely on reader funding to publish writing like this—and to keep our own paywall-free gift economy going. To support this work and read Harris’s essay in print, subscribe now and use code HARRIS20 at checkout for 20% off, just this week. For less than $4 a month, you’ll get our next four issues in the mail, starting with our latest—which also includes Zeynep Pamuk on the Trump administration’s war on science, Paul Hockenos on life in Russian-occupied Ukraine, Omer Bartov and Olúfẹ́mi O. Táíwò on memory politics in Israel and Brazil, Mark Bould on China Miéville’s new thousand-page novel, and more.
For further reading, don’t miss Alexander Hartley’s essay, “To Whom Does the World Belong?”—a deep dive on AI companies’ “original sin,” the mass appropriation of human-written text, and why we have to rethink intellectual property law—as well as Evgeny Morozov’s forum essay, “The AI We Deserve,” which insists on a “radically utopian” proposal: technology doesn’t have to serve markets and the military.
[](https://www.bostonreview.net/articles/knowledge-collapse/)
Knowledge Collapse
AI companies are racing to mechanize mathematics. Where does that leave human understanding?
- Michael Harris
- Alexander Hartley
- Evgeny Morozov with responses from
- Brian Eno
- Audrey Tang
- Terry Winograd
- Bruce Schneier & Nathan Sanders
- Sarah Myers West & Amba Kak
- Wendy Liu
- Edward Ongweso Jr.
- Brian Merchant
- X
- Bluesky
[](https://www.bostonreview.net/articles/to-whom-does-the-world-belong/)
To Whom Does the World Belong?
The battle over copyright in the age of ChatGPT.
[](https://www.bostonreview.net/forum/the-ai-we-deserve/)
The AI We Deserve
Critiques of artificial intelligence abound. Where’s the utopian vision for what it could be?
trending\_flat
###### Get our newsletter
Vital reading on politics, ideas, and culture to your inbox
This field is for validation purposes and should be left unchanged.
Email(Required)
---
A political and literary forum, independent and nonprofit since 1975
Registered 501(c)(3) organization
---
Help Us Stay Paywall-Free
Become a Member
---