Vitalik: AI May Accelerate Cryptanalysis, Recommends Prioritizing Hash-Based Encryption Schemes

Ethereum co-founder Vitalik Buterin warned in a post that AI-accelerated mathematical research could pose a severe threat to existing cryptographic systems within the next two years, particularly lattice-based cryptography (Lattices), ML-DSA, and FHE schemes, with ECDSA also facing a heightened risk of being compromised sooner than anticipated. Vitalik noted that AI could replicate breakthroughs akin to the transition from naive factorization to the Number Field Sieve, substantially undermining the security of lattice-based cryptography. In such a scenario, hash-based schemes would comprehensively outperform lattice-based ones in efficiency. The Ethereum Lean Roadmap is already progressing toward a "pure-hash" approach, employing signature schemes like WOTS or SPHINCS+ without incorporating any lattice-based components. From a personal standpoint, Vitalik offers the following recommendations: • Prioritize hash-based schemes, replacing lattice cryptography in feasible scenarios • Conservatively estimate lattice parameter sizes, with a recommendation to increase key sizes by a factor of 10 • Ensure privacy protocols avoid posting encrypted data on-chain, opting instead for off-chain transmission through third parties • Utilize offline confirmation for multi-signature wallets to prevent signatures from being publicly exposed • It is acceptable to retain funds in "fresh addresses" that have never initiated a transaction; however, address migration must be handled with extreme care, as losses resulting from operational errors may far exceed those caused by hacker breaches.