AI 4h ago 3 min read

Vitalik Buterin Warns AI Could Weaken Crypto Security in Two Years

Ethereum co-founder Vitalik Buterin warns AI-accelerated math research has a "good chance" of seriously weakening lattice-based cryptography within two years. He also flagged added ECDSA risk and said Ethereum's Lean roadmap is shifting toward hash-based signatures.

Key points

  • Ethereum co-founder Vitalik Buterin said on Thursday that AI-accelerated math research has a “good chance” of seriously weakening lattice-based cryptography, including ML-DSA and fully homomorphic encryption, within two years.
  • Buterin also flagged added risk for ECDSA, the signature scheme that Bitcoin and Ethereum wallets use, and said Ethereum’s Lean roadmap is moving toward hash-based signatures.
  • Ethereum Foundation researcher Justin Drake separately urged the industry to begin planning for “bunker mode,” a controlled migration of assets to addresses whose public keys stay hidden.

Ethereum co-founder Vitalik Buterin warns AI-accelerated math research has a “good chance” of seriously weakening lattice-based cryptography within two years, according to an X post published on Thursday. Buterin also flagged added risk for ECDSA, the signature scheme that Bitcoin and Ethereum wallets use to authorize transactions.

What Vitalik Buterin Warns AI Could Weaken

Lattice-based cryptography is a family of security methods built on hard geometry problems involving grids of points in many dimensions. Buterin named two lattice-based systems: ML-DSA, a digital signature scheme designed to resist quantum computers, and fully homomorphic encryption (FHE), a technique that lets computers process encrypted data without decrypting it. He challenged the common assumption summarized as “elliptic curves broken, hashes safe, lattices safe.”

Buterin compared the risk to factoring, where a mathematical shortcut called the general number field sieve (GNFS) forced encryption keys to grow over decades.

“What if there are skeletons in the closet like that, both for elliptic curves and lattices, that we are simply not smart enough to discover but bots soon will be?” Buterin wrote.

He said that if AI delivers 50 years of math in two years, that progress could plausibly include a GNFS-scale improvement in breaking lattices. Buterin described that scenario as “a very plausible world and something not at all extreme to predict.”

Ethereum Moves Toward Hash-Based Signatures

Buterin said Ethereum’s Lean roadmap has moved toward a hash-only direction, using signature schemes such as WOTS and SPHINCS+ in place of lattices. A hash function is a one-way mathematical function that turns data into a fixed-length fingerprint, and Buterin said hash-based methods lack the mathematical structure that such attacks would exploit. He added that hash-only approaches are already known to work for signatures and proofs, while public key encryption is harder because established mathematical results show it cannot be built from hashes alone.

For users, Buterin advised using fresh addresses where possible, gathering multisig signatures off-chain, avoiding on-chain encrypted notes for privacy, and using larger parameters for any lattice-based public key encryption. A multisig wallet requires several signers to approve a transaction. He also cautioned against rushed fund migrations, because the migration process carries its own risk of loss.

Ethereum Foundation researcher Justin Drake separately posted on X urging the industry to “calmly begin planning for ‘bunker mode.'” Drake recommended a controlled migration of assets to fresh addresses whose public keys remain hidden behind a hash, and he pointed to the 722 mathematical results OpenAI published on October 6.

Buterin framed the two-year window as a prediction and urged the industry to take the risks from AI-accelerated math seriously.

Editorial Note: Reported and edited by the Crypto India Magazine editorial team. We use AI tools to assist with research and drafting; every article is reviewed and fact-checked by our editors.

Interested in advertising with CIM? Talk to us!

Shailaja Pathak

Shailaja Pathak

Shailaja Pathak is a research scientist and a data driven analyst with a strong interest in technology. She enjoys writing and closely follows and covers the latest developments in AI and emerging tech in India and beyond.