Vitalik Buterin says AI risk does not justify rushed wallet moves

What happened

Vitalik Buterin said holders should not rush to move funds into new wallets, even as he warned that AI-assisted mathematics could weaken established cryptography sooner than expected.

His post answered a call from Ethereum researcher Justin Drake for the industry to prepare for bunker mode. Drake urged large holders to move funds to fresh addresses whose public keys stay hidden behind hashes.

Drake argued ECDSA could theoretically be broken in months rather than years, and said a breakthrough could happen within months in a worst-case scenario.

Buterin disputed the pace, not the risk. He wrote that users should not "scramble to move their funds to new wallets today," and said the industry should cut exposure to quantum-vulnerable and potentially AI-vulnerable cryptography.

Why it matters

Buterin's concern extends beyond quantum computing. He argued that humans may have missed mathematical shortcuts that AI systems could discover, and he said lattice-based systems such as ML-DSA and fully homomorphic encryption deserve the same scrutiny as elliptic curves.

He said lattice-based systems could take damage from AI math progress over the next two years, and asked what happens if flaws exist in both elliptic curves and lattices that humans have not found yet, but AI tools soon will.

Yehuda Lindell, head of cryptography at Coinbase, pushed back, saying there is "no evidence whatsoever" that decades-old assumptions behind elliptic-curve cryptography are close to failing.

For holders, rushing carries its own cost. Buterin said botched migrations had cost him more money than all hacks combined.

Background

Ethereum's long-term plan already leans toward hash-based signatures such as WOTS and SPHINCS rather than lattice methods. Buterin said hash-based methods lack the kind of mathematical structure that could be exploited by AI tools.

He noted that public key encryption is harder to build this way, and long-standing math results show it cannot rely on hashes alone. Public-key encryption is still required for communications, privacy networks, websites, and VPNs.

What is still unclear

• It is not settled whether AI will find weaknesses in elliptic curves, in lattices, or in neither.
• Drake's worst-case timeline of months is disputed. Lindell says the old assumptions are not close to failing.
• Whether hash-based tools can cover every use case remains open, because public-key encryption cannot rely on hashes alone.

Questions readers ask