Morning Minute: Ethereum Researcher Says AI May Break Crypto Encryption Before Quantum

What it is

Justin Drake (Ethereum Foundation researcher) posted the specific mechanism and timeline behind his "bunker mode" warning on October 7, 2026: AI-driven cryptanalysis recovering an ECDSA private key from an exposed public key in about a week using a large GPU cluster. The existing summary names the warning; the page supplies the mechanics of why it's plausible and what it would actually mean for existing holdings.

How it works

Every wallet has a private key and a derived public key. The public key stays hidden behind a hash until the wallet first sends a transaction — at that point the public key is permanently visible on-chain. Drake's concern is that an AI system could work backward from a visible public key to the private key. His basis: OpenAI published 722 new mathematical results on October 6, and Drake argues elliptic curves (the structure ECDSA relies on) have exploitable mathematical patterns that a sufficiently capable system could target, whereas the hash function protecting an unused public key is specifically designed to resist that kind of structural attack.

Preconditions / Affected

Only wallets that have ever sent a transaction are exposed, because that's what reveals the public key. Wallets that have only received funds remain protected behind the hash. Drake flags roughly 20,000 addresses believed tied to Bitcoin's creator — each holding 50 BTC with already-exposed public keys — as the likely first target if this capability materializes, calling them "Satoshi's shield" since an attacker would plausibly exhaust those before reaching smaller holders.

What to do

Drake's stated recommendation: begin a controlled, calm mass migration of self-custodied funds to fresh addresses that have never transacted (so the public key stays hashed/hidden), starting with the largest holders. He explicitly says don't panic and don't rush the migration itself. For holders not comfortable managing this, the article's stated alternatives are using ETFs, using a custodian such as Coinbase, or engaging experts if self-custody is retained.

Caveats

This is a single researcher's threat assessment, not a confirmed break — Drake frames the "months" timeline as a worst case to plan for, not a certainty. The cited evidence (OpenAI's 722 math results) is about general mathematical capability, not a demonstrated attack on ECDSA or elliptic curves specifically.