Ethereum researcher Justin Drake urges 'bunker mode' over possible AI attacks on wallet keys

Ethereum researcher Justin Drake urged crypto holders on Wednesday to prepare for “bunker mode” against a possible AI attack on wallet signatures, CoinDesk reported. In the worst case, he said, an AI model could find a shortcut through the cryptography protecting bitcoin and ether “in months, not years,” and he recommended large holders gradually move funds to fresh addresses.

A wallet's private key authorizes a payment, and the related public key lets the network verify the approval. Working backward from the public key to the private key is supposed to require an impractical amount of computing. Drake's concern is that AI could find a shortcut, letting an attacker recover the private key and spend the coins on ordinary computers, with no need for the quantum machines the industry has been preparing for.

No practical attack on bitcoin or Ethereum wallet keys has been demonstrated, and none appeared in the mathematics research CoinDesk reviewed. The warning followed OpenAI's release on Tuesday of 722 mathematical manuscripts produced by an unreleased model tested on roughly 4,000 research problems. Some of those findings carry computer-checkable proofs while others remain unverified and could contain errors, the company said.

Within a day, an outside researcher reran the computer check on one result — a new limit on how fast computers multiply large grids of numbers, a question mathematicians have worked on since 1969 — and found it held. The manuscripts came from the model OpenAI said last month had solved the Navier–Stokes problem, one of seven Millennium Prize challenges in mathematics; each result used on average computing power equivalent to roughly three hours of ChatGPT Pro reasoning, according to the company.

On Ethereum, any account that has ever sent a transaction has revealed its public key, and stablecoins and tokenized funds issued on the network are controlled through that same signature system. Earlier CoinDesk reporting puts millions of bitcoin in addresses whose public keys are visible onchain.

Drake's argument is that elliptic curves, the mathematics behind bitcoin and ether wallet signatures, follow tidy patterns a powerful enough model could learn to exploit, while hash functions scatter information so that little pattern remains. Issuer-side controls address a separate failure mode: Cardano's CIP-0113, which this publication covered, puts freeze and seizure rules inside a shared smart contract; those rules govern what an issuer or counterparty does, not what happens when a signature key is the target.