Mathematicians Call for OpenAI Boycott Over Flood of AI-Generated Proofs
The Association for Human Mathematics is urging a boycott of OpenAI after the company released more than 700 AI-generated mathematics manuscripts at once, with three papers retracted the next day due to a sign error. Fields Medalist Terence Tao warns that mass automated harvesting of open problems could leave entire branches of mathematics less fertile.
The Association for Human Mathematics has called for a boycott of OpenAI after the company released more than 700 AI-generated mathematics manuscripts in a single batch, an action that has unsettled researchers and prompted warnings about the long-term health of the discipline.
The scale of the release was unprecedented. Within a day, OpenAI was forced to retract three of the papers after a sign error was discovered, an embarrassing correction that critics say illustrates the risks of flooding a field with machine-generated results before they can be properly vetted.
Fields Medalist Terence Tao added his voice to the growing concern, warning that the mass automated harvesting of open problems could leave entire branches of mathematics less fertile. His caution points to a deeper anxiety: that AI systems, however capable, may exhaust the supply of tractable questions without contributing the kind of insight that human mathematicians build over years of careful work.
The Association for Human Mathematics, which advocates for the role of human judgment in mathematical research, framed the boycott as a defense of scholarly standards. The group argues that dumping hundreds of machine-written manuscripts into the literature at once bypasses the peer review and informal scrutiny that normally filter new claims. Even if most of the proofs are correct, the sheer volume makes it harder for researchers to identify which results are trustworthy and which deserve further study.
The retractions over a sign error have become a focal point for that argument. A sign error can invalidate an entire proof, and its presence in released manuscripts suggests that the AI-generated work was not subjected to rigorous checking before publication. For mathematicians, the episode raises questions about accountability: who is responsible when automated systems produce flawed results at scale, and how should the community respond when a single actor can reshape the landscape of an entire field overnight?
OpenAI has not issued a detailed public response to the boycott call. The company has previously positioned its models as tools to assist scientific discovery, but the episode highlights the tension between rapid deployment and the slow, deliberative pace that mathematics has traditionally relied on.
The debate also touches on how open problems function in mathematics. These unsolved questions give direction to research programs and attract the attention of graduate students and established scholars alike. If AI systems rapidly generate solutions or partial solutions to large numbers of them, the field could lose the organizing structure that those problems provide. Tao's warning about reduced fertility speaks to this concern: a branch of mathematics that has been stripped of its open questions may struggle to sustain the next generation of researchers.
Some mathematicians see the boycott as a necessary line in the sand, while others worry that refusing to engage with AI tools could leave the discipline behind as the technology improves. The Association for Human Mathematics has taken a clear position, but the broader community remains divided over how to integrate automated systems without undermining the norms that have made mathematics a cumulative and self-correcting enterprise.
For now, the retractions and the boycott call have forced a conversation that many researchers had been postponing. The question is no longer whether AI will play a role in mathematical discovery, but under what conditions its output can be trusted and how the field should protect the human practices that give mathematical knowledge its authority.
3
