Ethereum co-founder Vitalik Buterin said on October 8 that the cryptocurrency industry should take seriously the risk that AI-accelerated mathematics could weaken the cryptography protecting crypto wallets, while explicitly warning against rushing to move funds in response. "I don't recommend anyone scramble to move their funds to new wallets today," he wrote on social media, a message aimed at calming a debate his own comments helped ignite.
The warning followed a public call from Justin Drake, a researcher at the Ethereum Foundation, urging the industry to calmly begin planning for what he called "bunker mode." Drake argued that in a worst-case scenario, a breakthrough against ECDSA, the elliptic-curve signature scheme that secures most cryptocurrency wallets, could arrive within months rather than years, potentially before quantum computers reach the point of breaking such systems on their own. He pointed to OpenAI's publication of hundreds of AI-generated mathematical results this week as evidence that, in his words, "mathematical superintelligence is upon us."
Buterin, writing in response and in his own posts reported by The Block, said most of the industry's security planning has assumed that elliptic curves could eventually be broken while hash functions and lattice-based schemes stayed safe. He said he now believes there is a good chance lattice-based schemes, including the ML-DSA signature standard and fully homomorphic encryption, will "take serious hits" from AI-driven math over the next two years, and recommended the industry become more conservative about which lattice parameters it relies on and favor hash-based constructions where practical.
A Warning, Not a Panic
Buterin's specific advice for individuals was narrow: keep funds in wallet addresses that have never signed a transaction, since those addresses have not exposed the public key data a cryptographic break would need to exploit, and avoid posting encrypted notes onchain in favor of sending sensitive data through offchain channels. He also repeated that any migration should be deliberate. "It's very easy to lose funds from a misconfigured rushed upgrade," he wrote, adding simply: "don't rush anything."
The exchange has not gone unchallenged. Cryptographer Yehuda Lindell pushed back publicly, saying there is "no evidence whatsoever" that decades-old elliptic-curve assumptions have been broken by any AI system. The dispute echoes a wider unease among cryptography researchers this week, with one Decrypt report framing the central question as whether AI, rather than quantum computing, ends up being the technology that breaks today's encryption first.
No concrete demonstration of an AI system breaking elliptic-curve cryptography or a lattice scheme has been published, and Buterin's own estimates are framed as probabilities rather than established findings. For now, his message to the industry is to prepare contingency plans quietly rather than act on them, leaving open how long that calm can hold if more AI-driven mathematical results continue to arrive at the current pace.
