Buterin on AI-driven cryptanalysis: take it seriously, don't rush migrations

Serguey Asael Shinder10 October 2026, 09:472 min read

A debate about whether AI could break blockchain cryptography sooner than quantum computers ran through crypto circles this week. CryptoSlate's account, published on 8 October, sets out both sides.

The trigger. Ethereum researcher Justin Drake called for the industry to prepare for "bunker mode", including moving assets to fresh addresses whose public keys have never appeared on-chain. The context was OpenAI's publication on 6 October of new mathematical results produced by an internal model, some formalised in Lean. CryptoSlate notes that OpenAI reported no attack on blockchain cryptography, and neither Buterin nor Drake pointed to evidence that ECDSA has been practically broken.

Buterin's position, on 7 October: "I don't recommend anyone scramble to move their funds to new wallets today. But we should take the risks to cryptography from AI-accelerated math seriously." The substance of his argument:

• Structured mathematics is the exposure. Elliptic curves and lattice schemes rest on algebraic structures where better algorithms can appear — as the number field sieve once did against RSA. "If AI will bring us 50 years of math in 2 years," such advances could arrive faster than planned.
• That includes post-quantum lattices. He named ML-DSA, which NIST standardised in 2024, as an area where AI-driven discovery could erode security margins within about two years.
• Hashes are the hedge. Ethereum's "lean" roadmap has moved toward hash-based designs — signature schemes like WOTS and SPHINCS — and away from lattice signatures such as ML-DSA and Falcon.
• Practical steps without panic: keep funds where practical in addresses that have never signed a transaction (the public key stays hidden behind the address hash), and for multisig operators, gather signatures off-chain and rotate signer keys after use. He added that he has personally lost more money through failed migrations than through hacks.

Beyond blockchains. Public-key encryption — messaging, VPNs, Tor, TLS — cannot be built from hashes alone, so Buterin suggested those systems may eventually need much larger security parameters, possibly key sizes multiplied by up to ten.

What to do with it. For holders: reusing addresses was already a small exposure; avoiding it costs little. For custodians and wallet builders: the lesson is to make key rotation routine and tested now, so that a future migration — whether forced by quantum or by mathematics — is not the first one you run.

Source: CryptoSlate report on Buterin's response, 8 October 2026 — https://cryptoslate.com/vitalik-buterin-urges-calm-as-ai-raises-new-fears-over-bitcoin-and-ethereum-cryptography-security