Months, not years. That was the worst case Justin Drake, an Ethereum Foundation researcher, put on Wednesday for the signature scheme that guards most crypto wallets.
Drake’s argument, posted on X and reported by The Block and CoinDesk, is that AI-driven mathematics could break elliptic curve signatures before a quantum computer does. He defined a break as recovering a private key within about a week on available hardware.
His prompt was OpenAI’s release the day before of 722 mathematical manuscripts written by an unreleased model.
He urged the industry to plan calmly for what he called bunker mode: a controlled mass migration of funds to fresh addresses whose public keys have never appeared on a blockchain, with sophisticated holders moving first. He described it as a personal recommendation and did not claim anything is broken today.
Vitalik Buterin agreed the risk is real, then pushed back on the remedy. He told holders not to scramble, and said he has personally lost more money in botched migrations than in every hack combined.
The exposure is large. An address reveals its public key once it has signed a transaction, which covers every active Ethereum account. CoinDesk put the bitcoin sitting behind exposed keys at more than 6 million coins.
What did not arrive with the warning is an attack. No one has shown an AI-derived weakness in elliptic curves, and OpenAI’s papers do not contain one. Buterin added a harder problem. He expects lattice cryptography, the planned quantum-safe replacement, to take serious hits from the next two years of AI mathematics. If he is right, the industry is being told to leave one scheme without knowing which one to trust next.
