Ethereum researchers warn AI math could, at worst, break wallet signatures within months
Two leading Ethereum researchers warn that AI-assisted math could, in the worst case, break the signature system used by crypto wallets within months. The worst-case framing matters: it is a warning about a possible scenario, not a forecast, and the report says that so far no one has actually broken the current ECDSA signature scheme in practice.
Justin Drake called on the blockchain industry, in a post on X, to prepare a "bunker mode." His concrete advice is to move funds to addresses that have never signed a transaction. The reasoning is that signing reveals the public key, which an attacker could theoretically use to derive the private key and steal the funds. While only the hash of the public key is visible, the funds remain protected, because the key cannot be recovered from that hash.
Ethereum co-founder Vitalik Buterin agreed in a reply on X but warned against acting too fast. He said he has lost more money to botched migrations than to hacks. The article paraphrases his statements rather than quoting them.
Looking further ahead, Buterin argues for an architecture that relies as much as possible on hash functions alone. His reasoning is that even lattice-based schemes, often considered quantum-safe, could be weakened by advances in AI.
The piece is a secondary report from The Decoder on the X exchange. It does not name the specific AI math breakthroughs behind the debate or say which AI system or lab produced them.
Key facts
- Two leading Ethereum researchers warn AI-assisted math could, in the worst case, break the signature system used by crypto wallets within months.
- Justin Drake called in a post on X for a "bunker mode" and recommends moving funds to addresses that have never signed a transaction.
- Signing reveals the public key, which could theoretically let an attacker derive the private key; a visible hash of the key does not allow that.
- Vitalik Buterin agreed but warned against acting too fast, saying he has lost more money to botched migrations than to hacks.
- Long-term, Buterin favours an architecture relying as much as possible on hash functions alone, since even lattice-based schemes could be weakened by AI advances.
Why it matters
Wallet signatures are what protect crypto funds, so a credible threat to them would be serious. Here, two prominent Ethereum figures are openly discussing how quickly AI-assisted math could, in the worst case, undermine that system: within months. Buterin's long-term view goes further. He argues that even lattice-based schemes, often considered quantum-safe, could be weakened by AI advances, which is why he prefers designs built as far as possible on hash functions alone.
Who it affects
Holders of crypto wallets and the wider blockchain industry, which Drake asked to prepare a "bunker mode." The exposure he describes concerns addresses that have already signed a transaction, because signing reveals the public key. Addresses that have never signed show only the hash of the public key.
How to use it
Drake's practical recommendation is to move funds to addresses that have never signed a transaction. Buterin agrees with the concern but cautions against hurrying, noting that he has lost more money to botched migrations than to hacks. The source gives no step-by-step migration guidance beyond this.
How solid is it
This is a secondary report on posts made on X, and the source gives no date for them. Buterin's exact words are not quoted; the article paraphrases him. The source does not say what specific AI math breakthroughs prompted the debate, or which AI system or lab produced any of them. The "within months" timescale is framed as a worst case, not a prediction or a base case. No one has broken the current ECDSA scheme in practice so far.
Risks and caveats
The danger described is theoretical: an attacker could theoretically derive a private key from a revealed public key. The source does not say that wallets have been hacked or funds stolen through this attack. Rushing carries its own risk, as Buterin points out with botched migrations. His own long-term preference for hash-based designs is an argument about architecture, not a ready-made fix.
“bunker mode”
— Justin Drake, in a post on X, as reported by The Decoder