OpenAI reported on October 6 that an unreleased AI model generated 722 mathematical manuscripts covering 372 groups of results. The model attempted to solve approximately 4,000 previously unsolved problems, though these findings are currently under evaluation.
Cryptographer Matthew Green from Johns Hopkins University expressed concerns that AI could identify mathematical vulnerabilities in elliptic-curve cryptography, which secures Bitcoin and Ethereum. He noted that a reduction of 20 bits in security levels could make attacks roughly one million times cheaper.
The US National Institute of Standards and Technology approved post-quantum standards in 2024, including ML-KEM and ML-DSA. However, there is speculation that AI might also discover weaknesses in these lattice-based systems, potentially necessitating further industry adjustments.
Blockstream CEO Adam Back described the warnings as dubious. Blockstream has developed SHRINCS, a hash-based signature system using SHA-256 with signatures starting at 324 bytes, which has been tested on the Bitcoin Liquid sidechain.
Some market participants are responding to these concerns. Charles Edwards, founder of Capriole Investments, stated he has reduced his number of wallets and accounts and ceased using DeFi protocols as a precautionary measure.
