Ethereum researcher Justin Drake warned that artificial intelligence could break the Elliptic Curve Digital Signature Algorithm, or ECDSA, before quantum computers pose the direct threat many in crypto have been preparing for. He urged users to start moving funds gradually to fresh addresses.
Drake posted the warning on Wednesday, calling on the crypto community to shift into what he described as a "bunker mode." His concern was not centered on quantum hardware. It was tied to rapid progress in AI mathematics.
On X, Drake said recent AI developments, especially a series of public breakthroughs from OpenAI in mathematics, made it hard to ignore a scenario in which AI undermines the cryptographic foundation used by most cryptocurrency wallets today.
"The reasonable expectation now is that ECDSA may collapse before the quantum threat (qday), with the worst case being months, not years," Drake wrote.
Why Drake says AI progress changed the timeline
Drake pointed to two recent OpenAI disclosures. Last Tuesday, OpenAI said it had produced hundreds of new mathematical discoveries across algebra, theoretical computer science, and mathematical logic. Last month, the company also said that 10,000 autonomous AI agents working in parallel solved the Navier-Stokes equations in 88 hours, a problem described in the article as one of the most famous unsolved questions in fluid mechanics and one that had challenged mathematics for centuries.
Drake summed up the shift this way: "The last few days have brutally undermined human confidence in mathematical intuition. Assumptions long treated as self-evident are falling one by one."
Why he sees elliptic curves as exposed
Drake argued that elliptic curves may be especially vulnerable because their structure is mathematically rich. He contrasted them with hash functions, which are designed to remove algebraic structure as much as possible. Elliptic curves, by comparison, contain many mathematical properties that can be exploited, including the Schoof algorithm, the Frobenius map, and pairings.
His point was blunt: if human mathematicians have already found so many tricks, a much stronger reasoning system may find more.
ECDSA remains the signature scheme used by major chains including Bitcoin and Ethereum. If someone can derive a private key from a public key, the assets in that wallet can be stolen in full. Drake said the scenario long associated with Q-Day may no longer belong only to quantum computing.
His proposed response: controlled migration
Drake's personal recommendation was to begin what he called a controlled mass migration of assets. Large holders and high-net-worth accounts, he said, should move first because they have the most exposed value.
He also suggested that after each transaction signature, users should move the remaining funds to a new address.
The reasoning comes from how unused addresses work on Ethereum and Bitcoin. When a user creates a fresh wallet, the public key is hidden behind a hash, and only the address, or hash value, is visible. Once that address signs a transaction, the public key becomes exposed on-chain. If AI can ever work backward from a public key to a private key, every address that has already signed would become a target.
Drake also stressed that migration should be gradual. Moving too quickly could create operational mistakes or liquidity breaks. As he put it, "A rushed migration does more harm than good."
Q-Day may no longer be only about quantum hardware
Crypto discussions around Q-Day have usually focused on hardware timelines: when companies such as PsiQuantum, IBM, and Google might produce machines with millions of qubits. Drake's warning shifts the discussion toward the software side, where AI reasoning could arrive earlier.
The article noted that this is not the first time AI has been linked to cryptographic fragility. OP Labs last year published a 10-year post-quantum roadmap that explicitly called for phasing out ECDSA-based externally owned accounts by 2036 and moving to post-quantum signature systems. In March this year, Europol also published a report saying that the real target of quantum threats is not the blockchain protocol itself, but public keys already exposed on-chain, closely matching the risk scenario Drake described.
The key difference is speed. Quantum risk has usually been framed as a slow-moving 5-to-15-year problem. Drake said the AI window could be measured in months to years. If that assessment is right, the security assumptions behind crypto assets may need to be revisited much earlier than many had expected.