OpenAI's publication of solutions to hundreds of outstanding mathematical problems has prompted debate among mathematicians and raised concerns about the potential implications for internet security, cryptocurrencies and the verification of scientific research.
The release included full or partial solutions to 372 problems, documented across more than 700 research papers. The work was produced by a single experimental model using one prompt for each problem, according to the information provided.
The scale of the findings led computer scientist Scott Aaronson to dub the event the "Mathocalypse", calling it "surely one of the biggest days in mathematical history".
The discoveries cover algebra, geometry and differential equations, and include a breakthrough involving a modified version of the Riemann hypothesis.
The development follows an earlier effort in which 10,000 AI agents reportedly solved the 90-year-old Navier-Stokes equations, one of the Millennium Prize Problems, in under 90 hours.
Mathematicians raise concerns over rapid progress
The pace of AI development has astonished mathematicians, with some noting that models that struggled with GCSE-level mathematics two years ago are now operating at the level of leading mathematicians.
Professor Timothy Gowers compared the rapid progress to seeing a "tsunami" on the horizon.
The developments have also prompted ethical objections. Fields Medallist Terence Tao shared a statement from the Association for Human Mathematics calling on researchers to stop collaborating with OpenAI. The statement characterised the large-scale publication of papers as a "demonstration of power" rather than genuine scholarship.
An independent advisory board consulted by OpenAI also explicitly requested that the company stop testing advanced mathematical problems on proprietary internal models.
Concerns over internet encryption
The advances have raised questions about the security of public-key cryptography, which underpins secure web browsing, messaging and banking.
These systems rely on mathematical problems that are currently considered too difficult for computers to solve efficiently.
Cryptographer Matthew Green warned that accelerating AI capabilities in mathematics could eventually lead to algorithms capable of breaking public-key encryption.
There is no evidence that OpenAI's discoveries have broken current cryptographic standards. However, experts are concerned that AI could accelerate mathematical progress so rapidly that developers would have insufficient time to establish new defences.
Cryptocurrency industry urged to prepare
The potential risks extend to blockchain networks and cryptocurrencies, which rely on cryptographic systems to secure transactions and protect digital assets.
Ethereum researcher Justin Drake advised the blockchain industry to enter "bunker mode" in anticipation of mathematical superintelligence.
Ethereum co-founder Vitalik Buterin warned that rapid advances in AI mathematics could pose a more immediate threat to cryptocurrencies than quantum computers. He suggested that the technology could undermine lattice-based post-quantum cryptography unless key sizes were substantially increased.
Buterin also advised cryptocurrency holders to keep funds in unspent wallet addresses so that their public keys remained undisclosed to potential AI cracking algorithms.
Verification remains a challenge
Despite the scale of the reported discoveries, questions remain over the validity of the mathematical results.
Valerio Capraro, an associate professor, pointed out that human mathematicians had not yet verified whether OpenAI's claimed proofs were valid or correct.
Formal verification of a single research paper can take months, creating a potential bottleneck as AI systems produce mathematical claims faster than human experts can evaluate them.
The gap between the speed of AI-generated research and the capacity of mathematicians to verify it raises questions about how the scientific community will assess and understand future discoveries.
With inputs taken from Sky News, The Independent UK and the Washington Post.