Ethereum co-founder Vitalik Buterin has warned that artificial intelligence could accelerate mathematical advances enough to create new risks for cryptographic systems used across blockchains and the wider internet. In a detailed post, Buterin said the industry should take the possibility of AI-driven breakthroughs in mathematics seriously, while cautioning users against rushing into large-scale wallet migrations.
Buterin’s warning goes beyond the commonly discussed threat posed by quantum computing. He argued that researchers should consider not only quantum-vulnerable cryptography but also cryptographic systems that could become vulnerable if AI significantly accelerates the discovery of mathematical techniques capable of attacking their underlying structures.
The main areas he identified as concerning are ML-DSA, fully homomorphic encryption (FHE) and lattice-based cryptography. Buterin said the industry has traditionally viewed elliptic curves as vulnerable to future quantum attacks while assuming hashes and lattices were comparatively safe. He argued that assumption may need to be reconsidered.
According to Buterin, the potential problem is not necessarily that AI will immediately break existing lattice systems. Instead, AI could discover mathematical shortcuts that humans have not yet found, potentially reducing the amount of computational work required to attack problems considered secure today.
Why Buterin Is Concerned About Lattice Cryptography
Buterin compared the possibility to the historical development of integer factoring. He pointed to advances in algorithms such as the General Number Field Sieve, which dramatically improved the efficiency of factoring compared with more naive approaches and helped shape modern requirements for RSA keys and signatures.
The Ethereum co-founder questioned whether similar undiscovered mathematical shortcuts could exist for elliptic curves and lattices. In his view, AI systems capable of rapidly exploring mathematical problems could potentially uncover weaknesses that researchers have missed over decades of human research.
That concern is one reason Buterin said Ethereum’s roadmap has increasingly moved toward a “hash-only” approach for certain cryptographic applications. He said this direction avoids lattices, ML-DSA, Falcon and lattice-based commitments inside zero-knowledge proofs, with hash-based approaches such as WOTS and SPHINCS- being used for signatures and proofs.
Buterin acknowledged that hashes cannot solve every cryptographic problem. Public-key encryption, in particular, requires some form of mathematical structure, meaning systems need trapdoor mechanisms based on approaches such as group theory, lattices or code-based constructions.
He argued that this creates a difficult trade-off because mathematical structures can potentially contain exploitable properties. His personal view is that developers seeking long-term security should consider substantially larger parameters for lattice-based systems, suggesting that multiplying key sizes by 10 could be a plausible response in some scenarios.
The warning extends well beyond Ethereum. Buterin noted that public-key cryptography underpins secure communications, websites, anonymizing systems, VPNs and other internet infrastructure. A major breakthrough against widely used cryptographic assumptions could therefore affect digital systems far outside cryptocurrency.
For blockchain users, Buterin offered a more immediate recommendation: avoid unnecessary panic. He specifically said he does not recommend people scramble to move their funds to new wallets because of the warning.
At the same time, he said using previously unused addresses can provide an additional layer of protection if doing so is easy. He strongly cautioned users about migrations, however, noting that mistakes during wallet migrations can themselves result in financial losses.
Buterin also recommended that privacy protocols avoid putting encrypted notes directly onchain where possible. He suggested sending sensitive encrypted information offchain through third-party mechanisms instead, reducing the amount of potentially exposed cryptographic material stored permanently on public blockchains.
For multisig wallets, Buterin favors conducting confirmations offchain because this can prevent signer-wallet signatures from being exposed publicly. He argued that this approach could allow multisig systems to degrade more gracefully if ECDSA were unexpectedly compromised.
Buterin’s broader argument is that cryptographic security should be evaluated against more than today’s known attacks. His concern is that AI could accelerate decades of mathematical progress into a much shorter period, potentially forcing developers to reassess assumptions about elliptic curves, lattices and other structured mathematical systems.
For Ethereum and the wider cryptocurrency industry, the warning adds another dimension to the long-term security debate. Quantum computing remains a major reason for preparing new cryptographic standards, but Buterin argues that AI-driven mathematical breakthroughs deserve similar attention. His message is not that existing systems are already broken, but that developers should prepare for a world in which mathematical discoveries arrive far faster than current security models anticipate.
About The Author
