Vitalik Buterin: AI May Weaken Cryptography Within Two Years and Put ECDSA at Risk

Odaily News: Ethereum co-founder Vitalik Buterin posted on X that AI-accelerated mathematics research could pose new security threats to existing cryptographic systems, especially lattice-based cryptographic schemes, including ML-DSA and fully homomorphic encryption (FHE). He believes that AI progress in mathematics over the next two years could significantly weaken the practical security of lattice cryptography, and ECDSA may also face the risk of being broken sooner than expected.

Vitalik pointed out that over the past year, Ethereum's Lean roadmap has been gradually shifting toward pure hash-based cryptographic schemes, avoiding the use of lattice-based signatures and related zero-knowledge proof structures. He suggested prioritizing hash-based cryptography where feasible and adopting more conservative security parameters for lattice cryptography; privacy protocols should try to avoid storing encrypted information directly on-chain, and multisig wallets should prioritize off-chain signature confirmation.

In addition, Vitalik suggested that users, when operationally convenient, store funds in addresses that have not yet initiated transactions to reduce potential risk, but he does not recommend that users immediately migrate assets on a large scale. He emphasized that migration operations themselves also carry risks, and his own losses caused by migration mistakes even exceeded the losses brought by previous hacker attacks.