Ethereum co-founder Vitalik Buterin is urging caution over the possibility that AI-assisted mathematics could weaken crypto wallet security. Ethereum researcher Justin Drake has proposed preparing for “bunker mode”: gradually moving assets to fresh addresses whose public keys have not been exposed.
The concern is that mathematical discoveries could undermine wallet cryptography before sufficiently powerful quantum computers arrive. The timing and feasibility of that scenario have not been independently verified.
Buterin’s response combines support for taking that possibility seriously with a warning about the risks of moving money. “I don’t recommend anyone scramble to move their funds to new wallets today,” he wrote.
Why AI mathematics is prompting concern
Drake’s argument centers on the elliptic curve digital signature algorithm, or ECDSA, used to secure cryptocurrency wallets. He suggests that AI could uncover mathematical shortcuts that weaken the assumptions behind those signatures. His assessment does not establish a timetable for such a breakthrough.
In Drake’s view, the mathematical structure of elliptic curves may give advanced AI systems opportunities to discover attacks. He contrasts that structure with hash functions, which are designed to minimize it. This is his reasoning for favoring different cryptographic foundations; the proposed vulnerability remains hypothetical.
OpenAI’s mathematical research provides the backdrop. The company announced a collection of AI-generated mathematical results on October 6, spanning work produced by an internal model. It also said it was releasing computer-checkable formalizations of many proofs.
In a separate September announcement, OpenAI claimed that an internal system had produced a solution to the Navier–Stokes existence and smoothness problem. The company described roughly 10,000 concurrent agents and about 88 hours to reach the result, with formal verification taking additional time.
Those announcements concern mathematical research. The possibility of translating such progress into a practical attack on cryptocurrency signatures remains an extrapolation, and the timing is uncertain.
Quantum resistance may not settle the question
Buterin’s concern extends to lattice-based cryptography, including systems designed to withstand quantum attacks. He argues that AI-driven mathematical advances could reduce their security too, potentially forcing developers to reconsider the assumptions and parameters they use.
He suggested that lattice security could face significant challenges over the next two years. That forecast depends on future research producing advances relevant to those systems; it is not an established deadline for a cryptographic failure.
Buterin linked this concern to the hash-based direction of Ethereum’s lean roadmap. His preference reflects a judgment about which mathematical foundations may be less exposed to future discoveries. It does not establish that hash-based systems are immune to every possible advance.
The broader issue raised by the researchers is whether preparations for quantum computing also account for improvements in conventional mathematical attacks. Their warning asks developers to consider both possibilities when planning future security changes.
Fresh addresses come with migration risks
Drake’s proposed precaution is a controlled migration to addresses whose public keys remain concealed behind a hash. He suggested that large, technically sophisticated holders take the lead, and that remaining funds move to another fresh address after a transaction is signed. The proposal aims to limit exposure under his hypothetical attack scenario.
Buterin expressed conditional support for keeping funds in addresses that have not sent a transaction, where doing so is straightforward. The potential protection should be understood as a precaution, without treating it as a demonstrated defense against an AI attack.
Both researchers also emphasized the practical hazards of migration. Buterin warned about losses from mishandled transfers, while Drake cautioned that hurried migrations could cause more harm than good.
Their proposed response therefore depends on careful execution. The uncertainty surrounding future cryptographic attacks does not remove the immediate possibility of losing funds through a poorly managed move.