A bust-up between mathematicians over OpenAI's claim to have solved a famous maths problem is fuelling fears that artificial intelligence could undermine trust in the scientific process, with researchers also warning that AI-generated fiction is increasingly finding its way into journals.
OpenAI said last month that its AI agents had solved a longstanding problem relating to the Navier-Stokes equations in 88 hours. The body that runs a prize for whoever can solve the problem said it had "apparently been settled", but that its review would take time.
Hours before the announcement, Tristan Buckmaster, a mathematician at New York University, went public with concerns. Buckmaster had been working towards the same goal with Levent Alpöge, a mathematician employed by day by the rival company Anthropic, and the pair kept their drafts in OpenAI's coding tool, Codex. Buckmaster said his co-researcher was tipped off that news of their progress had "reached" OpenAI, though he did not specify how. He asked whether research stored in Codex could have fed into OpenAI's own work, while stressing he was "not accusing anyone of anything".
OpenAI said nobody at the company, human or AI, saw the other teams' work before it was published. It initially said that while unlikely, it could not "rule out" that anonymised data from use of its products may have helped improve its models. Days later, it said Buckmaster's Codex prompts from the two months before the announcement "could not have influenced the system in any way".
Beyond the dispute, the story highlights a growing issue about the use of AI in science: that it could undermine faith in the scientific process, raising questions about the novelty and accuracy of research. There is evidence that AI chatbots' tendency to hallucinate is increasingly appearing in science, and it is feared it will become difficult to separate true discovery from AI-generated fiction.
One threat is to scientific journals. Dalmeet Singh Chawla, a science journalist who has spent a decade reporting on research misconduct, said publishers' integrity staff would need to "drastically increase [in number and effort] if they're anywhere near the fight". "Right now, it seems like they're not," he said.
AI has also faced accusations of hype. Anthony Cheetham and Ram Seshadri, chemists at the University of California, wrote in the journal Chemistry of Materials that after going through a random sample of crystal structures predicted by a DeepMind model, they had "yet to find any strikingly novel compounds", arguing that many were "trivial adaptations of known materials". Their verdict was not that the method was wrong, but that the results need more scrutiny. DeepMind rejects the charge, saying it tested each compound for novelty.
OpenAI just claimed to disprove one of computer science's most fundamental assumptions. But why is this such a big deal? Here's our quick breakdown.
Imagine you have a bunch of switches, each either ON or OFF, and rules like:
Switch A must be ON, OR switch B must be OFF
Switch B must be ON, OR switch C must be ON
At least one of switches A, D, or F must be OFF
With 6 switches, there are only 64 possible combinations.
But with 100 switches, there are approximately 1.27 trillion trillion trillion combinations (2¹⁰⁰), and the rules can connect switches in complicated ways, making it extremely difficult verify which combinations work.
So the mathematical challenge is finding even ONE combination that satisfies every rule, or proving that none exists.
The assumption right now is that even with clever algorithms that eliminate huge numbers of possible combinations, solving the hardest 3-SAT problems still requires exponentially growing computation as you add more switches.
This type of problem is called 3-SAT, where each rule involves at most three variables.
And 3-SAT is NP-complete, meaning that if someone discovers a polynomial-time algorithm that solves every 3-SAT problem, they could theoretically solve every problem in NP efficiently.
So the Exponential Time Hypothesis (ETH) makes a stronger claim: even in the worst case, solving 3-SAT cannot be done in subexponential time.
But OpenAI claims to have found an algorithm that does exactly that, meaning the time required still grows extremely quickly, just fundamentally slower than ETH says is possible.
If the proof holds, it would overturn one of the foundational assumptions in theoretical computer science.
Read more about it:
The Navier-Stokes dispute tests whether research stored in a company's tools can be kept confidential, and who gets credit when AI aids discovery.
If AI-generated or hallucinated material slips into journals, the trust on which scientific research rests could erode.
AI is already central to celebrated discoveries such as AlphaFold's protein predictions, so the stakes of the integrity debate are high.
Background
AI is widely seen as a powerful tool for science, able to process far more data than the human brain can handle. Its poster child is Google DeepMind's AlphaFold, which in 2020 predicted 200 million protein folds in one year and earned its developers a Nobel Prize in Chemistry. But the same technology can produce confident falsehoods, and there is evidence this is increasingly happening in science. A computer scientist at Ghent University searched more than a million biomedical research papers for trigger words favoured by AI writing tools and found that by December last year, nearly 90% had AI fingerprints on them, though AI detection technology is not perfect.
Key Perspectives
Tristan Buckmaster: Raised concern that research he stored in OpenAI's Codex tool could have fed into OpenAI's own work on the Navier-Stokes problem, while stressing he was "not accusing anyone of anything".
OpenAI: Says nobody at the company, human or AI, saw the other teams' work before it was published, and later said the Codex prompts "could not have influenced the system in any way".
DeepMind: Rejects the charge that its predicted crystal structures lacked novelty, saying it tested each compound for novelty and dismissing unfamiliar structures for looking familiar misses the point.
Anthony Cheetham and Ram Seshadri: Found after sampling the supposed new crystals that they had "yet to find any strikingly novel compounds", calling many "trivial adaptations of known materials".
Dalmeet Singh Chawla: Says generative AI has "basically disrupted the whole scientific process" and that publishers' integrity staff will need to "drastically increase" their efforts to keep up.
What to Watch
The prize body's review of the Navier-Stokes solution, which it says will take time.
Whether OpenAI's explanation settles the dispute, and whether any further questions emerge about how information "reached" the company.
How scientific journals respond to the rise of AI-generated submissions and hallucinated citations in research.
12 claims in this story checked against 1 source: 12 supported.
Show each claim
SupportedA mathematician asked whether drafts stored in Codex could have fed into OpenAI’s maths work.BBC News: “He asked whether his research, stored in Codex, could have fed into OpenAI's own work”
SupportedThe story says concerns are growing that AI-generated fiction is finding its way into scientific journals.BBC News: “there is evidence that this is increasingly happening in science. Eventually, it's feared, it will become difficult to separate true scientific discovery from AI-generated fiction.”
SupportedThe dispute is fuelling fears that AI could undermine trust in the scientific process.BBC News: “that AI could undermine faith in the scientific process, raising questions about the novelty and accuracy of research.”
SupportedOpenAI said its AI agents had solved a longstanding Navier–Stokes problem in 88 hours.BBC News: “The company said its AI agents had cracked it in 88 hours.”
SupportedThe prize body said the problem had “apparently been settled”, but its review would take time.BBC News: “The body that runs a prize for whoever can solve the problem said it had "apparently been settled", but that its review would take time.”
SupportedBuckmaster and Alpöge had been working towards the same goal and kept their drafts in Codex.BBC News: “They had been working towards the same goal, and kept their drafts in OpenAI's coding tool, Codex.”
SupportedBuckmaster said his co-researcher was tipped off that news of their progress had “reached” OpenAI, without specifying how.BBC News: “Buckmaster said that his co-researcher was tipped off that news of their progress had "reached" OpenAI, though he did not specify how.”
SupportedBuckmaster stressed that he was “not accusing anyone of anything”.BBC News: “though he stressed he was "not accusing anyone of anything".”
SupportedOpenAI said nobody at the company, human or AI, saw the other teams’ work before publication.BBC News: “OpenAI said that nobody at the company, human or AI, saw the other teams' work before it was published.”
SupportedOpenAI initially said it could not rule out anonymised product-use data helping improve its models, then said Buckmaster’s Codex prompts could not have influenced the system.BBC News: “Initially it said that while unlikely, it could not "rule out" that anonymised data from use of its products may have helped improve its models. Days later, it said Buckmaster's Codex prompts from the two months before the announcement "could not have influenced the system in any way".”
SupportedDalmeet Singh Chawla said publishers’ integrity staff would need to “drastically increase [in number and effort] if they're anywhere near the fight”.BBC News: “He tells the BBC that publishers' integrity staff - the people who screen submissions for accuracy and reliability - will need to "drastically increase [in number and effort] if they're anywhere near the fight".”
SupportedCheetham and Seshadri said they had “yet to find any strikingly novel compounds” and that many were “trivial adaptations of known materials”.BBC News: “they concluded that they had "yet to find any strikingly novel compounds", arguing that many were "trivial adaptations of known materials".”
