Hoskinson Pushes Back on Buterin’s Warning About AI and Cryptography
Vitalik Buterin has warned that AI could help researchers find new mathematical attacks on cryptography. Cardano founder Charles Hoskinson pushed back, arguing that developers should prepare for future risks but not overhaul security parameters without a specific attack and evidence of its impact.
- Buterin raised a hypothetical risk, not a demonstrated attack.
- Hoskinson says lattice-based cryptography has faced decades of scrutiny.
- Coverage offers no public technical rationale for a reported proposal to multiply lattice key sizes by ten.
- AI-assisted research and quantum computing pose different potential threats.
A warning, not a reported break
Buterin’s concern is that increasingly capable AI systems could help mathematicians find more efficient ways to solve problems that cryptography relies on. If a breakthrough made an attack substantially more effective, developers might need to revisit existing security parameters.
Coinpedia reported that Buterin raised the concern on October 8 and Hoskinson responded within hours. The coverage describes a disagreement about how developers should prepare. It offers no evidence of a working AI attack on lattice cryptography or a broken algorithm.
According to Coinpedia, Buterin suggested lattice key sizes might eventually need to increase tenfold. The report does not say exactly which key measurements would grow or give a technical calculation for that figure. Hoskinson called the proposal “numerology, ” arguing that such a large change needs a defined attack and quantitative analysis. Hoskinson disputed Buterin’s warning.
That is a fair demand. There is no universal formula for translating an improvement in an attack into a fixed increase in key size. The impact depends on the algorithm, the attack’s cost, the desired security level and the practical burden of larger keys, including storage, bandwidth and verification costs.
What lattice cryptography relies on
Lattice-based cryptography depends on the difficulty of specific computational problems involving lattices, mathematical structures made up of regularly arranged points. Researchers have spent decades studying how to solve those problems more efficiently, including with lattice reduction and sieving techniques.
Hoskinson argues that researchers have accounted for known advances in attack methods when setting parameters for current lattice-based systems. But decades of scrutiny do not guarantee permanent safety. Cryptographic security rests on assumptions about which problems are computationally infeasible. New mathematics could change those assumptions.
The key question is whether a proposed advance is practical and how much it changes the cost of an attack. Comparing it with improvements in attacks on a different mathematical problem can help explain why cryptographers stay alert, but it does not show that lattice-based systems have the same weakness.
Post-quantum is not the same as AI-resistant
NIST has standardized ML-KEM for establishing shared cryptographic secrets under FIPS 203 and ML-DSA for digital signatures under FIPS 204. It has also standardized the hash-based signature scheme SLH-DSA under FIPS 205. NIST selected these systems for their expected resistance to attacks from sufficiently powerful quantum computers.
“Post-quantum” describes the threat these algorithms are designed to address. It does not mean they are immune to every possible mathematical discovery, including discoveries aided by AI. Standardization is no guarantee that a weakness will never be found. It reflects a selection process and a defined technical specification, not a promise of permanent security.
AI-assisted cryptanalysis and quantum computing are separate concerns. A quantum computer running Shor’s algorithm could threaten widely used public-key systems based on elliptic-curve cryptography. The concern about AI is different: it might help researchers discover or improve mathematical attacks. Quantum algorithms can also affect hash functions, but that requires a separate analysis and is not the same direct threat.
This debate looks ahead. It does not report that Ethereum or Cardano has moved current wallets to ML-KEM or ML-DSA, or that existing wallet signatures have been broken.
Hash-based cryptography has trade-offs too
Ethereum’s post-quantum research has considered hash-based signatures for uses that suit them. WOTS is a hash-based signature building block, while SPHINCS+ is a hash-based signature scheme. These approaches rely on different mathematical assumptions from lattice-based systems, making them options worth evaluating, but not automatic replacements for every cryptographic tool.
Hash-based systems still depend on secure hash functions, sound design and correct implementation. MD5 and SHA-1 have suffered collision attacks, but those weaknesses do not show that modern hash functions or hash-based signature schemes are broken. They do show why security claims must be assessed against the specific function and construction involved.
Different signature systems also have different performance and implementation demands. Blockchain developers must weigh key and signature sizes, network verification, wallet support and how safely users could migrate. No single cryptographic family suits every job.
Key questions and answers
-
Has AI broken lattice cryptography?
No working AI attack was reported in coverage of the dispute. Buterin raised a possible future risk, while Hoskinson challenged the case for major changes without evidence of a practical attack.
-
Are current Ethereum or Cardano wallets using ML-KEM or ML-DSA?
The exchange does not report either network deploying those NIST-standardized systems in current wallets. The debate is about preparing for possible future cryptographic threats.
-
Does post-quantum mean safe from every future attack?
No. It means an algorithm is designed to resist attacks from sufficiently powerful quantum computers. It does not guarantee protection against every future mathematical breakthrough.
-
Should users rush to move their funds?
The reported debate does not establish an immediate need for a mass migration. Rushed transfers can bring their own risks, and the coverage identifies no demonstrated break in the systems discussed.
Preparation needs a testable threat model
Buterin’s warning gives developers a reason to keep testing cryptographic assumptions before a crisis forces them to act. Hoskinson’s counterpoint matters too: precaution should not turn into arbitrary parameter inflation or a rush to replace one family of algorithms with another.
For blockchain teams, the practical work is to evaluate specific attacks, publish credible cost estimates and test migration plans before users need them. That beats declaring today’s cryptography invulnerable or treating a hypothetical breakthrough as proof that it has already failed.