Buterin: AI advances could undermine crypto cryptography in 2 years

Vitalik Buterin, co-founder of Ethereum, has cautioned that advances in AI-powered mathematics may weaken lattice-based cryptography within a two-year timeframe. He also indicated that the Elliptic Curve Digital Signature Algorithm (ECDSA), which protects Bitcoin and ETH wallets, could be compromised sooner than anticipated.

Despite the warning, Buterin advised against panic and said users should not hurry to transfer their funds now. He was responding to a post by Ethereum researcher Justin Drake, who has called for the industry to prepare a “bunker mode.”

Why Are Lattices Suddenly in Doubt?

Buterin noted that the common assumption has been that elliptic curves are vulnerable, while hashes and lattices remain secure. He now doubts the safety of lattices specifically.

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

The mathematics of lattices underpin the Module-Lattice-Based Digital Signature Algorithm (ML-DSA), a quantum-resistant signature standard. It also enables Fully Homomorphic Encryption (FHE), which allows computations on encrypted data.

Buterin referenced factoring, where the number field sieve method required encryption keys to enlarge over decades. He argued that if artificial intelligence can achieve 50 years of mathematical progress in just two years, lattices may encounter a comparable setback.

Hashes lack an exploitable internal structure, making them a more secure option in Buterin’s view compared to curves or lattices. Ethereum’s Lean roadmap, its long-term upgrade strategy, has been moving toward hash-only designs for about a year. The network’s signatures already employ hash-based methods like SPHINCS and WOTS.

In simpler terms, some of the latest security technologies depend on lattices—mathematical puzzles that quantum computers find hard to break. Buterin’s concern is that AI might uncover a shortcut to those puzzles, similar to past shortcuts that compelled older encryption keys to become longer. Hashes, which transform data into fixed-length codes with no exploitable patterns, appear more robust to him. This explains Ethereum’s shift toward hash-based signatures in its roadmap.

What About Bitcoin?

Bitcoin likewise depends on elliptic curve signatures, making Drake’s warning relevant to it as well. He stated that in a worst-case scenario, ECDSA could be compromised within months.

Drake highlighted Project11’s “risq list,” which tracks Bitcoin public keys already exposed on the blockchain. He noted that wallets with less than 50 BTC receive partial protection from what he called “Satoshi’s shield.”

The shield relates to about 20,000 exposed addresses linked to Bitcoin’s pseudonymous creator, Satoshi Nakamoto, each containing 50 BTC.

Drake added that Binance, Bitbank, Robinhood, Bitfinex, and Tether can seize an opportunity to strengthen their offline cold storage.

What Should Holders Do Now?

Buterin recommended that users keep funds in addresses that have never initiated a transaction, provided this is convenient. He noted that his personal losses from poorly executed migrations exceed his total losses from all hacks.

For multisig wallets that need multiple approvals, Buterin prefers offchain confirmations. Should ECDSA break, a multisig wallet would revert to being controlled by a single signer—the one gathering signatures—rather than being unlocked to anyone.

Regarding privacy protocols, Buterin strongly recommends transmitting encrypted notes offchain via a third-party channel instead of recording them on the blockchain. Drake added that Ethereum’s roadmap schedules need to be reviewed and sped up.

The implications extend beyond blockchains. Since public-key encryption cannot depend solely on hashes, secure messaging, VPNs, and website access all rely on structured mathematics. For any system intended to endure, Buterin recommends increasing key sizes by ten times.