Vitalik Buterin, creator of Ethereum, is urging the cryptocurrency industry to take seriously the threat that artificial intelligence could find mathematical methods capable of breaking the protections that prevent third parties from accessing money. crypto. Vitalik considers it necessary to prepare defenses, although he advises against moving assets or changing wallet settings hastily.
The Ethereum cofounder responds from his X account to Ethereum researcher Justin Drake, who yesterday, October 7, proposed starting to plan a bunker mode. Both raise preventive measures against future advances. Today, to move cryptocurrencies, the owner authorizes a transaction through a signature generated with their private key. The network verifies that it is valid using public information, without needing to know the secret that allows control of the funds.
Security depends on it being impractical to calculate that private key from the available information. However, Buterin fears that AI could discover mathematical shortcuts that drastically reduce the difficulty. If that happened, an attacker could end up signing transactions on the owner's behalf.
For his part, Drake believes the industry must anticipate that risk before there are quantum computers powerful enough to attack current systems. In his most adverse scenario, he suggests that ECDSA, a signature system used in Bitcoin and Ethereum, could be compromised within months.
What it means for those who hold cryptocurrencies
A hardware wallet keeps the keys isolated from the computer and allows transactions to be confirmed from the device. That protection helps against many cyberattacks, but it would not prevent a new mathematical method from obtaining the private key without accessing the device. Having assets deposited on an exchange or in a bank does not automatically eliminate this risk either. The entity can provide controls and manage updates, but, if it uses the same signature system, it depends on its mathematical foundations.
As a precaution, Drake proposes studying the use of addresses that have never signed transactions. In certain types of address, the public key remains hidden behind a cryptographic fingerprint until it is used. Keeping it unexposed reduces the information available to a hypothetical attacker, although the usefulness of this measure depends on the network and the type of address.
Buterin considers it reasonable when it can be easily applied, but warns that a mistaken transfer can cause an immediate loss. The developer acknowledges that he has lost more money to failed migrations than to all the attacks he has suffered combined.
The correction on multisignature wallets
In a later post, Buterin clarifies his proposal for wallets that require authorization from several people. The ideal rule, he explains, would be for each signer to change their key after every transaction. Thus, an exposed key would no longer serve to authorize future movements.
He also recommends gathering signatures off-chain when possible, to shorten the time between their publication and the withdrawal of the corresponding key. This defense means that obtaining the private key through an attack takes some time. If it were instantaneous, the attacker could try to get ahead of a transaction still pending confirmation. That is why Buterin presents rotation as a precaution under certain assumptions and again asks that no one make hasty or misconfigured changes.
Alternatives exist, but they require changes to the networks
Long-term protection requires incorporating other signature systems. Buying a more modern device is not enough, since the wallet must generate the new signature and the blockchain must recognize it.
In this regard, the United States National Institute of Standards and Technology, NIST, approved in August 2024 two post-quantum signature standards: ML-DSA, based on mathematical structures called lattices, and SLH-DSA, based on hash functions. The latter produce a digital fingerprint from the data and allow certain forms of authentication to be built.
However, Buterin prefers hash-based constructions where they can be used. In his view, AI could discover new ways to exploit the mathematical structure of lattices, forcing security parameters and transaction costs to be increased.
His warning does not prove that ML-DSA is broken. NIST keeps this algorithm among its standards and recommends its use. What Buterin questions is assuming that resisting quantum computing guarantees maintaining the same level of protection against future mathematical discoveries.
NEAR shows progress
For its part, the NEAR blockchain already allows its users to authorize transactions with ML-DSA, a signature system designed to resist attacks from quantum computers. However, that protection does not yet cover the entire network. NEAR must also update the systems used to verify transactions and add new blocks. Its goal is to complete that part by the end of 2027, although the timeline depends on research progress.
In addition, connections with other blockchains remain pending. Each network must accept the new signatures, and some upgrades do not yet have a date.