Ethereum co-founder Vitalik Buterin has warned that artificial intelligence could weaken key parts of cryptocurrency security within two years, including lattice-based cryptography that is widely viewed as resistant to quantum computing. Digital signatures are the mechanism by which users prove ownership of funds on blockchains, which is why warnings about the mathematics behind them matter to the entire industry.

Buterin issued the warning in a post on X on October 7, 2026 (X post). He said advances in mathematics accelerated by AI could expose weaknesses in cryptographic systems and urged the crypto industry to reduce its reliance on both quantum-vulnerable and potentially AI-vulnerable methods.

I don't recommend anyone scramble to move their funds to new wallets today. But 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.… — vitalik.eth (@VitalikButerin) October 7, 2026

Buterin Questions Common Security Assumptions

Buterin said the risk extends beyond older systems such as ECDSA — the elliptic-curve algorithm used to sign transactions on major blockchains — which is already considered vulnerable to quantum attacks. He said lattice-based methods, including ML-DSA and fully homomorphic encryption (FHE), could also face risks. ML-DSA is a post-quantum signature algorithm standardized by the US National Institute of Standards and Technology (NIST), while FHE enables computation directly on encrypted data, making both building blocks of current post-quantum and privacy research.

He challenged the common industry assumption that elliptic-curve cryptography will be broken while hash-based and lattice-based systems remain secure. Buterin said there is a “good chance” that the real-world security of lattice cryptography could suffer serious damage from AI-driven mathematical progress over the next two years.

He pointed to RSA, a decades-old public-key system that predates elliptic-curve methods, as an example of how security assumptions can carry hidden costs. According to Buterin, RSA keys and signatures need to be nearly 400 bytes rather than 64 bytes.

“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?” he wrote.

Buterin added that if AI produces “50 years of math in 2 years,” the resulting progress could include a major increase in the ability to break lattice-based systems.

Ethereum Favors Hash-Based Signatures

Buterin said Ethereum’s lean roadmap has focused on a “hash-only” direction rather than lattice-based cryptography. The approach includes signature types such as WOTS and SPHINCS.

Hash-based methods do not rely on a mathematical structure that attackers can easily exploit, he said, making them more resistant to AI-driven attacks. Hash-only approaches already work for signatures and proofs, although public-key encryption is more difficult because long-standing mathematical results indicate that it cannot be built from hashes alone.

For multisignature wallets, But said the ideal practice would be for each signer to change their key after every operation. He also recommended collecting signatures off-chain where possible, reducing the length of time that a key remains exposed.

He advised users to create fresh addresses wherever possible, avoid using on-chain encrypted notes for privacy, and select larger parameters for any lattice-based public-key encryption.

Other organizations are preparing for future cryptographic threats. Europol has urged the crypto industry to begin preparing for quantum-computing risks, warning that wallets are the primary point of exposure. The EU’s NIS Cooperation Group has called for a post-quantum strategy to be adopted by the end of 2026.

The Ethereum Foundation aims to make Ethereum’s main network fully quantum-resistant by 2029 — a target that sits outside the two-year window Buterin flagged for AI-driven cryptographic risk. The original report was published by Blockonomi.