Ethereum researcher urges crypto holders to harden wallet signatures against AI-driven breaks

A "bunker mode" call lands for bitcoin and ether holders

A researcher at the Ethereum Foundation, Justin Drake, publicly told crypto investors on October 8, 2026 to enter what he called "bunker mode" against possible AI attacks on wallet signatures. Multiple outlets reported Drake's framing that the signatures guarding those wallets could be broken "in months, not years" in the worst case, a timeline sharper than the years-long quantum-computer estimates practitioners have been working from . The call was specific in scope: ECDSA, the signature scheme behind standard bitcoin and ether addresses, was the scheme identified as the AI target .

For practitioners responsible for key custody, the change is the urgency: even if the worst case is unlikely, the warning came from inside one of the two largest protocol ecosystems, not from outside it.

Pushback from Coinbase and a warning against panic from Buterin

The warning drew immediate, named disagreement. Coinbase's chief cryptographer, Yehuda Lindell, called Drake's framing "the definition of FUD" and pushed back on the AI-key timeline publicly . Vitalik Buterin was reported to have warned holders against rushing to migrate, and Matthew Green and Dankrad Feist split from each other on the underlying AI threat . That pushback matters for workflow: it means the bunker-mode call is contested by people whose day job is the same cryptography, and not a settled consensus.

Why the AI-versus-quantum distinction matters for migration

A separate thread in the same day's coverage tied the AI warning back to a longer-running quantum conversation, noting that Vitalik Buterin has previously warned quantum computers could break bitcoin and Ethereum cryptography as soon as 2028 . The distinction matters to a practitioner planning a migration: an AI-driven math break on ECDSA and a fault-tolerant quantum computer running Shor's algorithm are two different threat models, with two different remediation paths and two different deadlines, and one of them is now being argued as potentially much closer than the other.

Where holders are stuck between two incompatible pressures

The practical consequence of the disagreement is that holders sit between two pressures. On one side, a researcher inside the Ethereum ecosystem is telling large holders to prepare for a possible AI-driven break of ECDSA within months . On the other, a Coinbase cryptographer and a founder of Ethereum are telling them not to panic-migrate on the same evidence . The result is a "do something, but don't do something drastic" position that leaves treasury, custody and exchange security teams with a hard call rather than a clear playbook.

What to watch next

Two concrete signals will tell readers which side of the debate the evidence is shifting toward. First, whether any independent group publishes reproducible work showing AI-assisted math breaking ECDSA on addresses with non-trivial value, rather than theoretical argument. Second, whether either Coinbase or the Ethereum Foundation ships a concrete migration tool, address-format change or custody recommendation in response, which would convert the public dispute into a deployable timeline. Until either happens, the bunker-mode warning is an internal argument between credentialed cryptographers, not an action item with a date attached.