Hoskinson and Buterin Clash Over AI Risks to Cryptographic Standards

The tension stems from Buterin’s suggestion that artificial intelligence might uncover mathematical shortcuts, potentially compressing decades of research into a brief window of discovery. He pointed to lattice-based cryptography—a cornerstone of modern security—as a point of failure, drawing parallels to historical advances in integer factorization. Buterin argued that if AI finds similar weaknesses in lattice structures, developers may need to increase key sizes by a factor of ten to maintain integrity.

Hoskinson dismissed this comparison as flawed, noting that the security of algorithms like ML-KEM and ML-DSA relies on over forty years of scrutiny, including the development of the Lenstra–Lenstra–Lovász algorithm. He characterized the call for larger keys as "numerology," arguing that security parameters should be dictated by the proven computational cost of potential attacks rather than speculative projections.

At the heart of the disagreement is the path forward for blockchain security. Ethereum researchers have increasingly leaned toward hash-based signatures, such as SPHINCS, suggesting they offer more robust long-term protection. Hoskinson countered that hash functions are not inherently safer, citing the historical failures of MD5 and SHA-1 as evidence that no cryptographic family is immune to discovery. He maintained that until a concrete attack is demonstrated and its cost established, abandoning standardized lattice-based methods in favor of unproven alternatives remains premature.