Ethereum researcher Justin Drake urges bunker mode planning over AI risk

What happened

Ethereum researcher Justin Drake is asking the crypto community to plan for what he calls "bunker mode". The idea is a controlled move of funds to new addresses whose public keys have never appeared on-chain.

Drake framed the plan as a way to limit fallout from a theoretical breakthrough that would expose private keys attached to visible public keys.

Large and sophisticated holders, in his view, should consider moving most assets to addresses that have never signed a transaction. After such an address sends funds, he recommends moving the remaining balance into another fresh address. He urged users to avoid a rushed response.

Drake tied his warning to OpenAI's recent developments in mathematical reasoning. OpenAI said the average result took the equivalent of about three hours of ChatGPT Pro reasoning, and that the model was tested on thousands of problems with many proofs published in computer-checkable Lean form.

The concern centers on ECDSA. Standard Ethereum externally owned accounts use ECDSA on the secp256k1 curve, and Bitcoin and Ethereum depend on elliptic-curve cryptography for ownership and transaction authorization.

Drake's worst case is that a hypothetical breakthrough would extract a private key from a public key in approximately 1 week using existing hardware. He said nothing suggests AI has broken ECDSA, RSA or the core cryptography of Bitcoin and Ethereum.

Why it matters

The risk comes from how keys leak. Once an account sends a transaction, its public key can be reconstructed from onchain data. Anyone who can efficiently derive the matching private key could take control of the assets.

The industry has traditionally treated that problem primarily as a future quantum-computing risk, and Drake challenges the idea that quantum computers must be the first technology able to break those systems.

He argues that sufficiently capable AI could uncover a classical algorithm that sharply reduces the difficulty of recovering private keys.

If that happened, the need to wait for large fault-tolerant quantum computers would disappear, which would change the industry's timetable.

What the data shows

OpenAI said the model was tested across thousands of problems and that the average result spent the equivalent of about three hours of ChatGPT Pro reasoning.

Drake's worst-case recovery time is approximately 1 week with existing hardware, and he asked whether 64-byte ECDSA signatures are too good to be true.

Background

Ethereum has already established a dedicated post-quantum security effort, with work underway on hash-based signatures, account abstraction and replacements for other cryptographic systems vulnerable to sufficiently powerful quantum computers.

Core post-quantum infrastructure is currently targeted for roughly 2029, although the roadmap remains subject to change.

Drake points to recent mathematical advancements that shattered long-time assumptions. He framed his proposal as prevention that can be done today with current wallets and crypto infrastructure, and insisted that at this stage it is a matter of preparation, not panic.

What is still unclear

• Whether a classical algorithm that breaks elliptic-curve cryptography exists remains unknown.
• At the moment, there is no publicly known attack that can extract a private key from a public key on the curves adopted in key cryptos.
• There is no evidence that AI has broken ECDSA, RSA or the crypto fundamentals of Bitcoin and Ethereum.

Questions readers ask