Vitalik Buterin Sounds Warning on AI as Ethereum Rethinks Network Security

Ethereum co-founder Vitalik Buterin flagged a risk to digital assets posed by artificial intelligence (AI) accelerated maths. This comes amid growing fears about future supercomputer vulnerabilities and recent drains on crypto projects.

AI Could Compromise Networks Within Years

In a recent X post, Buterin told the community that AI-accelerated mathematical advances could weaken lattice-based cryptography within two years. According to him, factoring takes a specific amount of time, but smart individuals optimized sieve numbers.

One solution would be to make RSA signatures and keys 400 bytes instead of 64 bytes. To neutralize threats, Ethereum adopted a lean roadmap focused on hash-only without ML-DSA, lattices, Falcon, etc.

“To me, that’s a very plausible world and something not at all extreme to predict. If AI will bring us 50 years of math in 2 years, then that 50 years of math may very plausibly include a “naive factoring -> GNFS” level of improvement to our ability to break lattices. In that world, lattices will still exist, but they will have to be significantly bigger to guarantee the same level of safety.”

While a hash-only approach can seem perfect for signatures and proofs, questions surround public-key encryption. Right now, crypto projects can use a structured trapdoor object. The dilemma is that, because it has structure, AI will likely try to exploit it.

Crypto users have also flagged concerns that AI development could negatively affect network structure. Despite scaling benefits, harmful tools in the hands of bad actors could expose the sector to widespread losses.

Ethereum researcher Justin Drake also wrote on X that AI will break digital asset wallets before quantum computers. He projects this will happen in months, unlike Buterin’s two-year estimate. To prevent losses, he suggested that holders rotate keys after signing transactions and move assets to unexposed addresses.

Previously, he stated that AI is becoming capable of hacking multiple systems. This follows growing fears about Q-Day, when supercomputers will potentially break digital asset cryptography.

Recent trials of new models have shown they can escape sandboxes and take down sites, but they stressed the underlying utility. The network uses models to verify protocols and application layers, identify bugs, and improve overall security.