An Ethereum Researcher Tells Crypto to Prepare for Bunker Mode

Ethereum researcher Justin Drake has told the crypto industry to start preparing for what he calls “bunker mode,” arguing that artificial intelligence could break the cryptography protecting blockchain wallets sooner than almost anyone planned for.

Drake posted the warning on X on Wednesday, days after OpenAI released hundreds of new mathematical findings. “Recent days have been humbling for human mathematical intuition. Long-held, unquestioned hypotheses have fallen,” he wrote in comments reported by Cointelegraph.

The target of his concern is ECDSA, the signature scheme that authorises transactions on Bitcoin, Ethereum and most other major chains. In a worst case, Drake said, it could fall within months rather than years.

What Drake Means by Breaking a Signature

Drake was specific about the threshold he is watching for. He defines a break as recovering someone’s private key in roughly a week using hardware that is already available to buy or rent, such as a large GPU cluster.

That bar matters. A theoretical weakness in a signature scheme is an academic paper. One a well-funded attacker can run on rented cloud machines inside a week is an operational problem for every exchange, custodian and holder.

Drake argues that elliptic curves are unusually structured objects. They come with a “rich structure, with room for fancy tricks like Schoof, Frobenius, pairings,” he noted, and structure is exactly what a mathematical attack feeds on.

How a Crypto Wallet Signature Actually Works

For anyone new to this, the mechanism is worth understanding, because it explains why Drake’s advice takes the shape it does.

Every crypto wallet holds a private key, which is a very large secret number. From that private key, the software derives a public key, and from the public key it derives the address that other people send funds to. Signing a transaction proves you hold the private key without ever revealing it.

The important detail is the order of exposure. An address is usually a hash of the public key, so the public key stays hidden until the wallet first signs a transaction. Once it has signed, the public key is published on-chain forever.

Breaking ECDSA would mean working backward from a public key to the private key. So a wallet that has never signed anything has one extra layer of protection: the attacker cannot see the number they would need to attack.

OpenAI’s Results Changed the Timeline

What moved Drake from theory to urgency was the pace of machine mathematics this year. In September, OpenAI said a system of roughly 10,000 coordinating AI agents had produced a solution to the 90-year-old Navier-Stokes problem in 88 hours, a claim first reported by CNBC and still being checked by mathematicians.

Whether or not the proof survives peer review, it suggests long-standing problems can now be attacked at a scale no human research group can match.

Cryptography has always rested on an assumption that certain problems stay hard. Drake’s point is that the assumption was based on human progress rates, and those may no longer be relevant.

Vitalik Buterin Agrees on the Risk, and Warns Against Panic

Ethereum co-founder Vitalik Buterin publicly backed the underlying concern. “But we should take the risks to cryptography from AI-accelerated math seriously,” he wrote, while cautioning holders not to rush.

Buterin has pointed to his own losses from failed migrations, and a panicked scramble to move funds would cause more immediate damage than a hypothetical future attack.

Drake made the same point himself, warning that “a rushed migration would do more harm than good.” Dragonfly managing partner Haseeb Qureshi was among others weighing in on the debate.

Fresh Addresses and Hash-Based Signatures

The practical advice is to migrate slowly and deliberately rather than in an emergency. Drake suggests moving funds to addresses that have never signed a transaction, so the public key stays behind its hash, and moving again after any signature.

For institutions, he suggests auditing which public keys are already exposed in cold storage, and considering rotating keys for critical signers or adding hash-based signatures alongside existing ones.

Longer term, he favours cryptography built on hash functions, which have far less internal structure for a clever attack to exploit, and accelerating Ethereum’s planned security upgrades. The groundwork already exists elsewhere: in August, StarkWare demonstrated the first quantum-resistant Bitcoin transaction in block 964,199, though at a cost of $150 to $200 per transaction.

Could AI Arrive Before Q-Day?

The industry has spent years planning for “Q-Day,” the moment a quantum computer becomes powerful enough to break current signatures. Most roadmaps assume that is a problem for the 2030s.

Drake argues the ordering may be wrong, and that an AI-assisted mathematical break could come first. If it does, the defences being built for quantum computers may simply be needed earlier than scheduled.

Markets have not reacted to any of this. Bitcoin’s price sat near $82,700 on Thursday, down around 1.7% on the day and at month-to-date lows, with traders focused on bond yields and Middle East tensions rather than signature schemes.

Cryptography Becomes a Moving Target

Nothing has been broken. No funds have moved because of an AI attack, and Drake is describing a scenario, not an event.

What has changed is the shape of the assumption. Crypto was built on the idea that certain mathematical problems would stay hard for decades, letting the industry treat cryptographic upgrades as a slow background project rather than a live operational task. If machine-generated mathematics keeps advancing at this year’s rate, that project may need a schedule.

The debate now running through Ethereum’s research community, and across the latest crypto news, is less about whether the threat is real and more about how fast an industry holding trillions of dollars can safely change its locks.