Vitalik Buterin posted on X that there's a "good chance" AI-accelerated math could seriously weaken lattice-based cryptography within two years, specifically schemes like ML-DSA and fully homomorphic encryption, the kind many systems were counting on as quantum-resistant.
He's recommending hash-based signatures instead, naming SPHINCS+ and WOTS as alternatives, and telling people not to scramble and move funds immediately.
Here's what's actually new here versus what isn't.
Ethereum's own roadmap has been quietly shifting toward hash-based signature schemes for about a year already, something tied to the broader direction Ethereum's been heading. This AI warning isn't introducing a brand-new plan. It's adding a more urgent justification to a shift that was already underway.
Worth being precise about what the claim actually rests on too.
Buterin's logic is that AI could compress roughly 50 years of mathematical progress into two years. If that happens, lattice-based systems would need much larger parameters to stay secure, large enough that hash-based alternatives would simply become more efficient anyway.
That's a reasoned, probabilistic argument. Not a demonstrated exploit. None of the outlets covering this identified an actual technical mechanism by which AI would break the math, just Buterin's own "good chance" framing of a forecast.
I think that distinction matters.
"AI could weaken this within two years" and "AI has found a way to weaken this" are different claims. This is clearly the first one, a risk assessment from someone with real authority on the topic, not a confirmed vulnerability.
There's a small irony worth noting too.
Buterin's been a consistent, public supporter of AI tools generally, and here he's warning about AI as the threat to the exact cryptographic foundations crypto depends on. Using AI optimism and flagging AI risk aren't contradictory exactly, but they're two sides of the same person's thinking happening at once.
I'm not dismissing the warning. If the underlying logic holds, the stakes are real.
Just noting this is a forecast backed by reasoning, not an announcement of a working attack, and the shift it's reinforcing was already in motion before this post went out.
