OpenAI Math Breakthroughs Trigger Cryptography Fears Across Crypto Industry

OpenAI Publishes Hundreds of AI-Produced Math Results

OpenAI has released progress on more than 300 math research problems, with the company saying its internal model produced 372 math breakthroughs, nearly all from a single prompt to one AI agent. According to coverage in Scientific American, some results may have taken multiple attempts. The model was tried on about 8,000 problems and solved roughly 5% of longstanding open mathematical problems after a single 3-hour attempt on each, according to a post by Scott Aaronson, computer science chair at UT Austin. In total, OpenAI published 722 mathematical results, including a GitHub repository with mathematical manuscripts and supporting proof artifacts.

Researchers Warn of Cryptography Risk

Aaronson wrote that AI companies have started, in his words, gingerly and discreetly investigating whether their latest internal models can break important cryptographic protocols and primitives. He noted the striking under-representation of cryptographic breakthroughs among OpenAI's published results. Vitalik Buterin said the risks to cryptography from AI-accelerated math should be taken seriously and that exposure to potentially AI-vulnerable cryptography should be minimized, while adding that he does not recommend anyone scramble to move funds to new wallets today. Ethereum researcher Justin Drake called on the blockchain industry to calmly begin planning for what he termed bunker mode, and Ethereum researchers said users should consider moving funds to fresh addresses. Bitcoin slipped below $83,000 as the debate divided the crypto industry. Cryptographer Yehuda Lindell responded that there is no evidence pointing to a break of decades-old hardness assumptions such as elliptic curve cryptography.

Mathematicians Push Back on OpenAI

The Association for Human Mathematics said OpenAI's new math documents show power, not scholarship, and urged mathematicians to stop working with the company. The group said mathematicians did not ask for this work to be done and accused OpenAI of a total disregard for the norms of scientific research. Fields Medalist Terence Tao reposted the statement. The response drew sharp criticism on social media, with commentators calling the statement gatekeeping and comparing mathematicians to a medieval guild.

Funding and Model Launches

In other news, Butterfly Effect, the parent company of Manus, raised more than $500 million led by Boyu Capital and IDG Capital, in its first funding round since Beijing forced Meta to unwind its $2 billion acquisition of the company. Separately, Anthropic launched Claude Haiku 5.5, described as the first Haiku model with effort controls, aimed at high-volume, cost-sensitive tasks such as summary and classification requests. Anthropic called it the cheapest, fastest, and most capable small model it has released.

Sources: Techmeme