OpenAI says its artificial intelligence has made major advances in mathematics, physics, and computer science. But extraordinary discoveries require extraordinary verification. As independent researchers examine the findings, a bigger question is emerging: Who determines whether an AI-generated scientific breakthrough is genuinely correct?

WHAT’S HAPPENING

Artificial intelligence is moving beyond answering scientific questions and beginning to challenge problems that have resisted some of the world’s leading researchers for decades.

On October 26, 2026, OpenAI announced three major research results generated by its internal AI models.

The findings involve mathematical geometry, theoretical physics, and computational complexity—fields that underpin scientific research and modern technology.

The first concerns the Kakeya conjecture, a longstanding mathematical problem involving the smallest possible space needed to accommodate a line segment pointing in every direction.

OpenAI says its AI proved a related mathematical result that would resolve the conjecture across all dimensions.

The second involves Yang–Mills theory, which helps describe fundamental forces governing subatomic particles.

OpenAI claims its AI established an important mathematical existence result. However, the complete Yang–Mills Millennium Prize Problem remains unresolved because it also requires proving the existence of a positive mass gap.

The third—and potentially most disruptive—concerns the Exponential Time Hypothesis (ETH).

This widely studied computer science conjecture concerns how the difficulty of certain computational problems increases as their size grows.

OpenAI claims its AI produced a mathematical argument disproving ETH.

Rival AI company Anthropic announced an independent result challenging the same hypothesis.

Now researchers are examining whether these claims withstand scientific scrutiny.

The central development is not simply that AI generated remarkable mathematical arguments. It is that researchers must determine whether those arguments establish the discoveries being claimed.

WHY IT MATTERS

Scientific breakthroughs traditionally depend on researchers developing arguments that other experts can examine, reproduce, and challenge.

AI is introducing a new complication.

Advanced models can generate enormous amounts of mathematical reasoning, sometimes at a speed that overwhelms the researchers responsible for evaluating it.

OpenAI has used a mathematical verification system called Lean to formally check portions of its work.

Lean can verify whether a formal proof follows logically from its stated assumptions.

However, formal verification does not automatically establish that the mathematical statement being checked accurately captures the original scientific problem.

Independent experts must still evaluate its meaning and implications.

On November 19, the University of California, San Diego confirmed that computer scientists Russell Impagliazzo, Mohan Paturi, and MIT’s Ryan Williams were facilitating an independent examination of the ETH findings.

Their involvement provides an important avenue for scientific scrutiny, but it does not constitute automatic endorsement of the claims.

The controversy also raises questions about transparency.

OpenAI’s Advisory Group on Mathematics and Artificial Intelligence has called for broader public access to AI-generated mathematical work so researchers can evaluate important claims independently.

WHO BENEFITS

Mathematicians and scientists could gain powerful tools for investigating problems that previously required years of research.

Universities and research institutions may accelerate discoveries through collaboration between human experts and AI systems.

Technology companies could eventually benefit from improvements in algorithms, computing efficiency, and scientific modeling.

Students and researchers may gain access to new methods for exploring complicated mathematical ideas.

WHO LOSES

Independent researchers without comparable computing resources could find themselves disadvantaged as AI companies generate discoveries at unprecedented speed.

Scientific journals and academic institutions may struggle to evaluate increasingly large volumes of AI-generated research.

Researchers whose work is not independently verified risk building further investigations on incorrect assumptions.

The public could be misled when preliminary scientific claims are presented as established breakthroughs.

WHAT HAPPENS NEXT

The immediate priority is independent verification.

Researchers must determine whether OpenAI’s and Anthropic’s mathematical arguments are correct, whether their assumptions are appropriate, and what their conclusions actually establish.

Even if the ETH findings are confirmed, they would not automatically solve the famous P versus NP problem or demonstrate that today’s encryption systems have been defeated.

Likewise, the claimed Yang–Mills existence result would not qualify as a complete solution to the million-dollar Millennium Prize Problem.

The next major development will be whether independent mathematicians and computer scientists confirm the results and establish how they affect existing research.

The larger signal: Artificial intelligence may be entering a new era as a contributor to scientific discovery. But generating a proof and establishing scientific truth are not necessarily the same achievement. As AI becomes capable of producing increasingly sophisticated research, the responsibility for verification, transparency, and scientific accountability becomes more important—not less. The future of AI-driven discovery may depend as much on humanity’s ability to validate new knowledge as on machines’ ability to generate it.

Stay Sharp

Subscribe to follow the Trend newsletter and more.

Have a tip or idea?

Pass along insights or story ideas on AI, startups, and business. Focused on signal over noise, impact over headlines. Facts. Trends. Consequences. Always.