Ethereum researchers urge calm preparation as AI math raises fresh questions for wallet signatures

What happened

Ethereum researchers used posts on Wednesday to argue that progress in artificial intelligence and mathematics should change how the industry prepares for a future break of the signature scheme that secures most crypto wallets. The scheme is ECDSA, short for elliptic curve digital signature algorithm. It lets a holder prove control of an address by signing with a private key while publishing only a signature and, in many cases, eventually exposing the related public key.

The initial warning called for the industry to begin planning what was described as bunker mode. In practical terms, that meant a controlled migration of funds toward fresh addresses whose public keys have not been exposed. Large and sophisticated holders were urged to move first. The same recommendation included moving remaining funds to a new address after signing a transaction, because a signed transaction can reveal information that a fresh receiving address does not.

The response from Ethereum's co-founder backed the underlying concern while rejecting panic. He said he does not recommend that anyone scramble to move funds to new wallets today, but added that the industry should take risks to cryptography from AI-accelerated math seriously. Keeping funds in addresses that have not yet been used to make a transaction was described as a good precaution when the move is easy. He also cautioned that a botched migration can cause losses, saying he has personally lost more in failed migrations than in hacks.

The trigger for the exchange was a release of new mathematical work. Reporting on the posts said OpenAI released hundreds of mathematical findings on Tuesday across fields including algebra, theoretical computer science and mathematical logic. The same reporting recalled that last month the company used 10,000 autonomous AI agents working in parallel to solve a famous fluid dynamics problem in 88 hours. One researcher described recent results as humbling for human mathematical intuition and argued that elliptic curves could be especially open to attack because they carry rich mathematical structure. In a worst-case framing, that researcher said a breakthrough against ECDSA could arrive in months rather than years, before quantum computers reach the point usually called Q-day.

No working break of ECDSA was reported. The posts describe a risk scenario and a preparation plan, not evidence that current wallets can be drained by AI today.

Why it matters

ECDSA sits under a large share of crypto custody. Bitcoin, Ethereum and many other networks use elliptic curve signatures to authorize spending. If a practical method emerged to recover a private key from an exposed public key within a useful time frame, addresses that have revealed their public keys would face a different risk from addresses that have only received funds. That distinction is why the proposed preparation focuses on address hygiene rather than on selling assets or changing networks.

The AI angle changes the timetable discussion. Most long-term planning for signature risk has been built around quantum computing. A large quantum computer running the right algorithm is a known theoretical threat to elliptic curve systems, and networks have debated post-quantum signature schemes with that event in mind. The researchers are arguing that conventional mathematics, accelerated by AI systems that can search for structure and test ideas at scale, could weaken assumptions sooner or in a different order than the quantum plan assumes.

That argument reaches beyond ECDSA. The co-founder's response said the common planning mode has been to treat elliptic curves as broken in a future attack, hashes as safe and lattice-based systems as safe. He warned that the concrete security of lattices could take serious hits from the next two years of AI math, naming ML-DSA and fully homomorphic encryption among the areas at risk. Lattice-based schemes are a leading family of post-quantum candidates. If AI advances reduce their safety margin at the same time ECDSA is under pressure, a migration plan built on one replacement family could prove less durable than expected.

The response also pointed to a design direction for Ethereum. He said the possibility is a major reason the network's lean roadmap has been moving toward hash-only constructions where possible. Hash-based signatures have a different structure from elliptic curves and lattices, with security tied more directly to the strength of the hash function. For privacy protocols, he advised against putting encrypted notes onchain and in favor of sending them offchain through another mechanism. That is a product design point as much as a cryptography point: data placed permanently on a public ledger can be revisited later with better tools.

The strongest practical point in the exchange may be the warning against haste. Moving funds sounds simple until it involves multisignature wallets, exchange deposit addresses, estate arrangements, tax records, smart contract allowances and institutional approval chains. A rushed migration can send funds to a wrong address, strand assets behind a lost key, break accounting records or expose a holder to phishing at the moment they are most distracted. Both researchers stressed a gradual, controlled process for that reason. Preparation here means inventory, procedures and tested wallets before any emergency, not a same-day scramble after a headline.

What to watch

Watch for evidence, not volume of posts. A credible claim in this area would include a reproducible method, a clear statement of cost and time, and independent review by cryptographers who did not produce the result. A faster factoring or discrete logarithm result on small test curves would be a signal to study. It would not by itself prove that production wallet keys are at risk.

Watch what wallet and custody providers do. If major providers begin offering fresh-address rotation tools, clearer warnings when an address has exposed its public key, or supported migration paths for multisignature and institutional wallets, that would show the warning is moving from research discussion into product work. If providers stay silent, the practical effect for most holders will remain limited.

Watch protocol roadmaps for how they treat signature replacement. Ethereum's stated lean toward hash-based constructions, and any parallel work in Bitcoin and other networks, will show whether developers are preparing a second path if ECDSA weakens faster than expected. Parameter choices matter as well. The co-founder argued for more conservative lattice parameters where lattices are still used. Future proposals that raise those margins, or avoid lattice assumptions in new designs, would be a concrete response.

Finally, watch the split between large holders and everyday users. The recommendation puts large and sophisticated holders first because they face the highest value at risk and have the staff to move carefully. A disorderly rush by small holders, following social media summaries rather than provider guidance, would create the migration mistakes both researchers warned against. The measured reading of Wednesday posts is simple: no ECDSA break has been shown, AI math is advancing fast enough to change preparation timelines, and any move to fresh addresses should be planned, tested and unhurried.