Vitalik Buterin has raised concerns that advances in artificial intelligence could put the cryptography behind Bitcoin and Ethereum at risk.
He believes AI could speed up mathematical research so much that progress, which might normally take more than 50 years, could happen in around five.
One of the problems he mentioned is the elliptic curve discrete logarithm problem (ECDLP). It plays a key role in protecting private keys on both networks. If researchers found an efficient way to solve it, an attacker could potentially derive a private key from its corresponding public key and gain access to the funds.
No practical method for doing this is publicly known. Buterin’s concern is that AI could help researchers find a mathematical shortcut that has so far remained undiscovered.
He pointed to the quadratic sieve as an example of how a new algorithm can change the assumptions behind cryptographic security. Developed in the early 1980s, it made factoring large numbers much faster and affected the security assumptions used by RSA encryption. There is no known equivalent shortcut for ECDLP, but Buterin believes AI could accelerate the search for one.
Buterin also mentioned pairing-based cryptography, which is used in Ethereum’s validator signature system. Validators rely on cryptographic signatures to take part in the network’s consensus process. He also raised concerns about lattice-based methods, which form the basis of several leading post-quantum cryptography standards.
The reference to post-quantum cryptography brings another issue into the discussion. These standards are being developed to protect systems against future quantum computers. Buterin’s point is that AI could help uncover weaknesses in the mathematics behind cryptographic systems, including methods designed to withstand quantum attacks. He did not say that any of the existing standards had already been broken.
The topic of AI security has also appeared in Buterin’s recent writing. In October, he discussed a “paranoid mode” for private chats and supported the development of an AI model deliberately trained with reduced hacking capabilities.
Buterin raised the cryptography concern on November 3, the opening day of Devcon 8 in Mumbai. His keynote also touched on trustworthy AI, privacy and secure technology.
His warning does not mean that Bitcoin or Ethereum has been compromised. There is no confirmed breakthrough showing that ECDLP can now be solved efficiently, and no evidence that the cryptographic systems used by either network have been broken.
For now, the concern is about what AI could make possible in the coming years. Buterin’s estimate of five years is a warning about the potential speed of mathematical progress, not a prediction that Bitcoin and Ethereum will be hacked within that timeframe.
