OpenAI withdraws three preprints a day after releasing 722 manuscripts on unsolved math problems
OpenAI pulled three AI-generated mathematics manuscripts from a public repository just one day after publishing a massive dump of 722 papers on unsolved problems. The initial October 6 release detailed progress across geometry, computer science, algebra, and other fields from an unreleased internal model. The retractions occurred after a sign error invalidated a key proof, which subsequently broke two dependent papers related to the Hodge conjecture. Alongside the retractions, the company revised 14 other manuscripts and updated 13 citations, while separate analyses noted translation errors in a Navier-Stokes proof.
Listen to Live Briefing
Real-time synthesized voice briefing · Live Feeds Desk
- ✓ OpenAI published 722 preprints in a GitHub repository on October 6 describing progress on 372 math problems.
- ✓ The company withdrew three manuscripts less than 24 hours after release due to a sign error.
- ✓ The sign error invalidated a key proof and impacted two dependent papers related to the Hodge conjecture.
- ✓ OpenAI revised 14 other manuscripts and updated 13 citations.
What changed
OpenAI withdrew three preprints and revised 14 others less than 24 hours after publishing 722 AI-authored math manuscripts.
Live updates
-
OpenAI Withdraws Three Math Preprints Following AI Proof Errors
OpenAI pulled three AI-generated mathematics manuscripts from a public repository just one day after publishing a massive dump of 722 papers on unsolved problems. The initial October 6 release detailed progress across geometry, computer science, algebra, and other fields from an unreleased internal model. The retractions occurred after a sign error invalidated a key proof, which subsequently broke two dependent papers related to the Hodge conjecture. Alongside the retractions, the company revised 14 other manuscripts and updated 13 citations, while separate analyses noted translation errors in a Navier-Stokes proof.
Why it matters
The mass release of artificial intelligence-authored proofs triggered intense reactions across the academic mathematics community, with observers describing the initial dump using terms ranging from breathtaking to devastating. This incident highlights the fragility of automated mathematical reasoning and exposes the severe risks of errors propagating through dependent research. Rigorous human verification remains critical as artificial intelligence systems generate complex manuscripts at scale.
What is confirmed
- OpenAI published 722 preprints in a GitHub repository on October 6 describing progress on 372 math problems.
- The company withdrew three manuscripts less than 24 hours after release due to a sign error.
- The sign error invalidated a key proof and impacted two dependent papers related to the Hodge conjecture.
- OpenAI revised 14 other manuscripts and updated 13 citations.
Still unconfirmed
- About 42% of the released manuscripts are Lean-checked.
What to watch next
- Further independent audits and peer review of the remaining 719 AI-generated math manuscripts.
- Official technical disclosures from OpenAI regarding the unreleased model used to generate the proofs.
confidence 100%Sources used for this update (13)
- WSJ — AI Solved One Math Problem and Everyone Freaked Out. It Just Cracked Hundreds More.
- New Scientist — The most interesting mathematical discoveries in OpenAI’s 722 new papers
- The New York Times — ‘Breathtaking,’ ‘Devastating’: Mathematics Reels After New OpenAI Release
- The Conversation — Is this the ‘mathocalypse’? Why OpenAI’s latest results dump has left mathematicians in shock
- Retraction Watch — OpenAI withdraws three preprints a day after releasing 722 manuscripts on unsolved math problems
- New Scientist — OpenAI mistranslated mathematics into code for its Navier-Stokes proof
- retractionwatch.com — OpenAI withdraws three preprints a day after releasing 722 ...
- www.online-tech-tips.com — OpenAI Math Release: What’s Real in the 722 AI Manuscripts
- www.europesays.com — OpenAI withdraws three preprints a day after releasing 722 ...
- www.aipulse.it — OpenAI Withdraws Three Math Preprints After Sign Error in AI ...
- rits.shanghai.nyu.edu — OpenAI Drops 722 Math Manuscripts From an Unreleased Model
- aiweekly.co — OpenAI drops 722 AI math preprints; three pulled over sign ...
Community Sentiment: How do you assess this situation?
Voice your perspective · Real-time aggregated sentiment from the Live Feeds community