Editorial illustration for OpenAI Forms Advisory Group Amid Concerns Over Centralized Power
OpenAI's Math Advisory Group Tackles AI Power Concerns
OpenAI Forms Advisory Group Amid Concerns Over Centralized Power
OpenAI brought about 40 mathematicians together in August to talk through a problem the company itself created: what happens when its models start solving math faster than the field can process it. Attendees left with a mix of excitement and dread, according to Northwestern University's Bryna Kra, who was in the room. The company suggested its systems had already cracked hundreds of longstanding problems and asked the group for guidance on how to release that work responsibly.
Mathematicians pushed back on the idea of dropping results in a blog post or a tweet, the way OpenAI had done with ten problems earlier that same month. Kra says the group specifically asked for papers, something mathematicians could actually read, check, and build on, rather than a dump of claims with no explanation. Now OpenAI appears set to publish hundreds of those solutions on GitHub, according to people familiar with the plans, raising the same concerns Kra says got ignored the first time.
The episode has become a test case for how a company holding increasingly capable AI systems handles a scientific community it doesn't fully control.
“On August 28, we began training a new internal model. In addition to resolving the Navier—Stokes Millennium Prize problem, this model has now resolved more than 100 long-standing open problems across most areas of mathematics,” says McCallum.
Why this matters
The math community's reaction is a preview of what every research field will face once a single company can plausibly claim to have solved hundreds of problems before anyone else even attempts them. OpenAI convening 40 mathematicians, then walking back an assurance about staggered releases, tells us the company itself isn't sure how to handle the asymmetry it's created. For researchers, that's the real story: if OpenAI's models can quietly generate solutions faster than peer review can process them, the normal incentives for working on hard problems start to erode.
The advisory group announced in mid-September is OpenAI's attempt to manage that fallout, but it's an advisory group with no enforcement power, chosen and convened by the same company it's meant to check. The mathematician's line about "accumulation of power in one place" applies well beyond math. Anyone building on OpenAI's models, or competing with them, should watch whether this group gets real say over release decisions, or whether it's just a sounding board that lets OpenAI say it listened.
Related Reading
Browse all OpenAI news →
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
Open Source Box CEO Aaron Levie Says AI Will Drive "Biggest Cybersecurity Upgrade
Open Source Study: 6.7% of Deepfake Requests on Hugging Face Targeted Children
Related Topic OpenAI researcher quits, citing distrust over ad‑driven engagement metrics
LLMs & Generative AI
Related Topic OpenAI launches GPT-Image 1.5 with precise editing for enterprise visuals
LLMs & Generative AI
From Archives Mistral Claims ‘Le Chonk’ Is Top Open-Weight AI Model Outside China
From Archives Telcos Build AI Strategy on Open Models for Local Adaptation
From Archives OpenAI Adds Invisible Watermark to ChatGPT and Codex for EU Users
LLMs & Generative AI
From Archives Wikimedia Links May Outage to 'Rogue' OpenAI Bots
Business & Startups
Common Questions Answered
What specific mathematical breakthrough did OpenAI's new internal model achieve in August?
OpenAI's internal model resolved the Navier-Stokes Millennium Prize problem, which is one of the most challenging unsolved problems in mathematics. Additionally, the model has resolved more than 100 long-standing open problems across most areas of mathematics since training began on August 28.
Why did OpenAI convene an advisory group of 40 mathematicians?
OpenAI gathered 40 mathematicians to discuss how to responsibly release mathematical solutions that its models have already solved, recognizing that the company's systems are solving problems faster than the field can process them. The company sought guidance from the mathematical community on managing the responsible disclosure of these breakthroughs.
What concerns does the math community have about OpenAI's centralized power in research?
Mathematicians are concerned about the asymmetry created when a single company can solve hundreds of research problems before other researchers even attempt them, potentially disrupting the traditional peer review and collaborative research process. The math community's mixed reaction of excitement and dread reflects worries that OpenAI's capabilities could fundamentally alter how research is conducted and validated across the field.
What does OpenAI's handling of mathematical solutions reveal about the company's preparedness?
OpenAI's decision to convene mathematicians and then walk back assurances about staggered releases suggests the company itself is uncertain about how to manage the power asymmetry it has created in the research landscape. This uncertainty indicates that OpenAI lacks a clear strategy for responsibly releasing breakthrough solutions in a way that benefits the broader research community.
Further Reading
- Advisory Group on Mathematics and Artificial Intelligence - OpenAI
- Top mathematicians will advise OpenAI on releasing its AI-generated mathematical results - The Next Web
- OpenAI says its internal model solved over 100 long-standing math problems after just a month of training - The Decoder
- OpenAI's model solves 100+ maths problems, company forms advisory group - Business Standard
- Nine mathematicians will advise on releasing OpenAI's AI-generated math results - Cryptopolitan
