Will AI Break Crypto Encryption? Ethereum’s Vitalik Buterin Weighs In on 'Bunker Mode' Shift

What it is

Vitalik Buterin (Ethereum co-founder) posted on X (Oct 7, 2026) that it is "plausible" AI-accelerated mathematics could break lattice-based cryptography within two years — a risk distinct from quantum computing. This follows Ethereum Foundation researcher Justin Drake's call a day earlier for "bunker mode": moving funds to addresses whose public keys have never been exposed on-chain.

How it works

Lattice-based schemes (used by projects like Cardano for quantum-resistance) are considered quantum-safe but are not proven safe against classical mathematical breakthroughs. Buterin's argument: if AI compresses 50 years of math progress into two years, that progress plausibly includes attacks on lattice problems specifically, since lattice security relies on open mathematical conjectures rather than hardness assumptions as battle-tested as ECC/RSA factoring. A wallet's public key is only exposed once it has made an outgoing transaction (it's revealed to broadcast the signature); addresses that have only received funds and never sent have not exposed their public key, so an attacker who breaks the underlying crypto still has nothing to attack.

Preconditions / Affected

• Relevant to any address whose public key has already been exposed via a prior outgoing transaction — ECC/RSA addresses are already quantum-vulnerable in the long run, but the new concern is specifically lattice-based schemes thought to be quantum-safe.
• Projects evaluating or deployed with lattice-based post-quantum cryptography (e.g., Cardano's considered approach) are the ones with near-term exposure if Buterin's two-year estimate holds.
• Does not apply to addresses that have never transacted outward (public key never exposed).

What to do

• Buterin explicitly says not to scramble and move funds today, but to "take the risks... seriously" and minimize exposure to both quantum-vulnerable and potentially AI-vulnerable cryptography going forward.
• His stated preparation step: keep/move funds into addresses that have never made an outgoing transaction, since their public keys remain unexposed.
• He now favors hash-based cryptography over lattice-based schemes for quantum-proofing, while noting hash-based schemes could theoretically also be weakened by future AI-driven math (no specific timeline given for that).
• If deploying or evaluating lattice-based PQC, Buterin suggests increasing lattice sizes as a mitigation.

Caveats

• This is a probability estimate ("plausible," "good chance") from one researcher's public post, not a demonstrated break — the article notes "the post doesn't demonstrate that these systems have been broken" (Ananda Banerjee, Charlie Quant Lab).
• Banerjee explicitly cautions against treating the two-year figure as a deadline and recommends careful preparation over panic/rushed fund migration.
• One month prior, Buterin had argued AI would not doom crypto security and that defenses would keep pace — this post marks a reversal in his public stance, not a new technical disclosure.