Vitalik Buterin has warned that artificial intelligence could accelerate the discovery of mathematical vulnerabilities in lattice-based cryptography, potentially compromising schemes like ML-DSA and fully homomorphic encryption (FHE) sooner than anticipated.
In a post on X, Buterin outlined a core concern: while many in the crypto community consider elliptic curves vulnerable and hashes and lattices secure, there is significant risk that lattice-based systems will face serious security degradation from AI-accelerated mathematical research within the next two years.
The Mathematical Precedent
Buterin drew a parallel to historical developments in factorization. Naive factoring algorithms require 2^(n/2) time, but researchers later developed number field sieves that reduced this to 2^O(n^(1/3)), necessitating RSA keys to grow from 64 bytes to approximately 400 bytes for equivalent security.
He raised the possibility that similar "skeletons in the closet" may exist for elliptic curves and lattices. While humans may lack the capability to discover such weaknesses, Buterin suggested that AI systems soon will.
Ethereum's Shift to Hash-Based Designs
Ethereum's roadmap has moved toward hash-only cryptographic constructions over the past year, excluding lattices, ML-DSA, Falcon, and lattice-based commitments in zero-knowledge proofs. Signatures now rely on hash-based schemes such as WOTS and SPHINCS-.
Buterin noted that signatures and proofs can already transition to hash-only designs. However, public-key encryption presents a greater challenge, as mathematical theorems show it cannot be constructed from hashes alone and requires a trapdoor object with usable structure, such as lattices or code-based systems.
Recommended Security Measures
Buterin advised multiplying public-key encryption key sizes by 10 for long-term security against potential AI-driven attacks. He does not currently recommend increasing the byte size of hashes, suggesting that if concerns intensify, round counts should be increased first.
On practical fund management, Buterin advised against rushing to migrate funds to new wallets, citing personal experience with losses from botched migrations exceeding losses from hacks. He recommended keeping funds in addresses that have not yet conducted a transaction where practical, and favored offchain multisig confirmations and offchain delivery of encrypted privacy notes.
