Justin Drake Urges “Bunker Mode” for Ethereum (ETH): AI Could Break ECDSA Keys in a Week

Update 14:25 UTC: Drake's original post, published on X on October 7, spells out the mechanics of the call: a "controlled mass migration of assets to fresh addresses", meaning addresses whose public keys stay hidden behind a hash.

Justin Drake's “Bunker Mode” Call

Ethereum (ETH) researcher Justin Drake has asked the sector to start planning a defense he calls “bunker mode” against a cryptography risk he describes as real in direction but unproven in timing: as of Wednesday, October 7, no known attack breaks the signature schemes securing Ethereum or Bitcoin. The Ethereum price slipped about 4.6% over the past 24 hours, though the warning addresses long-term key security rather than any market event. Drake's proposal is specific. Large, technically capable holders should evaluate moving funds, in a controlled way, to new addresses whose public keys have never appeared on-chain, and once an address signs its first transaction, any remaining balance could be shifted again to a fresh address. He framed the step as preparation for a possible future break, not a response to an existing one, and said it can be scheduled gradually with today's wallets and custody infrastructure. The concern centers on ECDSA, the Elliptic Curve Digital Signature Algorithm, the standard method most networks use to confirm that a transaction was approved by the holder of its key. Transaction signatures across virtually every smart contract platform rest on the same elliptic curve assumptions, which is why he treats the issue as a sector-wide problem for the entire Ethereum ecosystem. The driver behind his timing argument is speed: recent progress in mathematical reasoning at OpenAI suggests, in Drake's view, a coming “mathematical superintelligence” phase in which machines find methods human mathematicians miss. He notes that several long-standing mathematical assumptions have recently been overturned, and argues a comparable acceleration in cryptography cannot be ruled out.

Q-Day, GPU Clusters and a One-Week Break

Drake argues the industry must take seriously the possibility that ECDSA could be weakened before the widely expected “Q-Day”, the point at which quantum computers break public-key cryptography. His worst case: a large GPU cluster, the kind of hardware available today, could derive a private key from an exposed public key in roughly a week. He also points out that algorithms built for quantum machines have, in some cases, later inspired classical ones, so a Shor-like method for elliptic curves and RSA running on ordinary computers could theoretically emerge. Elliptic curve cryptography, he argues, is structurally more exposed to advanced AI systems because of its algebraic structure, while hash functions are deliberately designed to carry as little of it as possible. On the defense side, Drake wants large crypto companies and institutional custody providers to lead. He named Binance, Bitbank, Robinhood, Bitfinex and Tether as firms that could strengthen their cold wallet structures, a discipline that extends, in his framing, to sizeable Ethereum treasury positions held in cold storage. He said Bitcoin addresses whose public keys are already visible on-chain deserve particular monitoring, while small Bitcoin holders may have relatively more time in some cases. He flagged thousands of addresses attributed to Satoshi Nakamoto, each holding about 50 BTC, as potential priority targets for attackers, calling the cluster “Satoshi's shield”. Infrastructure that signs continuously, such as oracle systems and layer-2 security councils, should evaluate rotating its ECDSA public keys on a regular cycle, with hash-based schemes like SPHINCS as an additional layer. Ethereum's own roadmap already includes a move to hash-based cryptography and end-to-end formal verification for the proof-of-stake network, but Drake says recent AI developments justify re-examining and accelerating that timeline.

The AI evidence Drake cites now has concrete form. On Tuesday, OpenAI published 722 mathematical results from an unreleased internal model, organized into 372 families in a public GitHub repository, with most proofs formalized in Lean so machines can verify them. The model was given roughly 4,000 problems, and Drake responded that "mathematical superintelligence is upon us." Precedent for AI finding cryptographic weaknesses already exists in one case: in July, Anthropic reported that its Claude Mythos Preview improved the best-known attack on HAWK, a post-quantum signature candidate submitted to NIST, cutting its key strength by 50% within 60 hours. The company said HAWK is not deployed, so production systems were unaffected. Separately, a Europol report this week counted 6.04 million BTC, about 30.2% of supply, with exposed public keys as of May 2026.

The warning has drawn support from the top of the project. Ethereum (ETH) co-founder Vitalik Buterin said he does not recommend anyone scramble to move funds today, but argued the risks to cryptography from AI-accelerated math deserve serious weight, extending the concern beyond ECDSA to lattice-based systems such as ML-DSA, FHE and Falcon. He said there is a good chance lattice security takes hits from the next two years of AI math, a reason Ethereum's lean roadmap has moved toward hash-only signatures using WOTS or SPHINCS, and floated 10x key sizes as one defensible choice for long-term security. He also cautioned that migrations carry their own losses, saying botched migrations have cost him more than all hacks combined, and advised off-chain approvals for multisig wallets so a broken ECDSA degrades security only to a 1-of-1 state.

Drake's remarks as reported on X on October 8 sharpened his timeline and added a caution. He said preparing for ECDSA to fall before Q-Day is reasonable and that, in a worst case, the break could arrive within months rather than years. At the same time, he warned holders against rushing: a hasty large-scale migration could itself cause greater losses, and hash-based addresses are not immune to every attack vector either. He also tied the acceleration case directly to Ethereum's long-term technical roadmap, Strawmap, which adopts hash-based cryptography and formal verification against the quantum threat, arguing that schedule should be pulled forward given the pace of AI-driven mathematical discovery.

Drake's original post, published on X on October 7, spells out the mechanics of the call: a "controlled mass migration of assets to fresh addresses", meaning addresses whose public keys stay hidden behind a hash. Early reaction in the market commentary has not matched his request for calm, and commentators point to alternatives for holders unwilling to manage the migration themselves, including ETF wrappers or an established custodian such as Coinbase. The market backdrop is uneasy: Bitcoin ETFs recorded $485 million in net outflows on Wednesday, their largest single day since late June, while Ethereum (ETH) ETFs saw $161 million leave, as crypto majors fell another 2% to 3% alongside stocks.

No Break on Record

The load-bearing record here is Drake's own framing, and it is consistent across both parts of his remarks: a directional risk with an undated arrival. What is established is the operational half: fresh-address hygiene, key rotation for signing-heavy infrastructure and hash-based fallbacks are all executable with today's tooling. What is not established is any demonstrated break. No researcher has produced evidence that ECDSA, RSA or the cryptographic foundations of Ethereum or Bitcoin have been compromised, and Drake himself labels the one-week key-recovery figure a worst case rather than an observation. Until someone shows a working method, bunker mode remains a planning exercise, and that is exactly how its author presents it.