Vitalik Buterin, co-founder of Ethereum, has called on the cryptocurrency community to take potential threats from artificial intelligence to cryptographic systems seriously. He specifically highlighted the growing risk to lattice-based schemes such as ML-DSA and fully homomorphic encryption (FHE) amid rapid advances in mathematical research fueled by AI.
AI and the new threat to cryptography
Buterin addressed these concerns in a detailed post on X. He suggested that while many in the crypto space consider elliptic curves broken and assume hashes and lattices remain secure, this confidence may be misplaced. Lattice-based cryptography, which has been considered post-quantum secure, could soon face serious challenges from AI-accelerated mathematical discoveries.
Buterin advised, “We should take the risks to cryptography from AI-accelerated math seriously and minimize our exposure to not just quantum-vulnerable cryptography, but also potentially AI-vulnerable cryptography.”
He compared the current state of cryptography to the evolution of factoring. Initially, basic factoring methods required exponential time, but later advances such as the number field sieve dramatically reduced this complexity. As a result, cryptographic systems like RSA required significantly larger keys than originally anticipated.
Buterin warned, “What if there are skeletons in the closet like that, both for elliptic curves and lattices? Humans may not be smart enough to discover them, but bots soon will be.” ML-DSA, FHE, and other lattice-based cryptographic tools were cited as especially vulnerable in this context.
He outlined a scenario in which artificial intelligence could achieve the equivalent of fifty years of mathematical progress in just two years, necessitating far larger lattice parameters. In such cases, hash-based alternatives would surpass lattice constructions in efficiency wherever applicable.
Mini dictionary: Fully Homomorphic Encryption (FHE) allows computation on encrypted data without needing to decrypt it first, enabling privacy-preserving operations in cloud computing and other sensitive environments.
Ethereum’s approach and Buterin’s recommendations
Over the past year, Ethereum’s lean roadmap has shifted toward hash-based cryptographic schemes and moved away from lattice-based systems such as ML-DSA, Falcon, and related commitments within zero-knowledge proofs. Ethereum now prioritizes hash-only digital signatures, such as the Winternitz One-Time Signature (WOTS) and SPHINCS- schemes, for efficiency and security.
Buterin acknowledged that while signatures and proofs can already rely on hash-based methods, public-key encryption remains challenging. Current theorems indicate that encryption cannot be achieved with hashes alone but requires a “trapdoor” structure, a fundamental property provided by lattices or code-based cryptography.
He suggested that AI-driven mathematical advances could undermine these trapdoor structures. To strengthen long-term security, Buterin recommended multiplying key sizes by a factor of ten.
He noted, “Hash-based [cryptography] is better than lattice-based, in those situations where hash-based is possible at all.”
Buterin clarified that he does not see an immediate need to increase the byte size of hash-based systems, but he would increase the number of rounds if concerns rise. As for extreme cases, he said only a dramatic discovery such as P = NP would break hash security, calling such an event very unlikely.
Advice for users and wallet management
Buterin recommended practical measures for individual users. He advised keeping funds in addresses that have never made a transaction, whenever possible, to reduce exposure.
He warned of the risks associated with hastily migrating wallets, stating that he had personally lost more money due to errors in migration than through hacks. For sensitive operations, he favored off-chain multisig confirmations and delivering encrypted privacy notes off-chain to add further protection.
