Charles Hoskinson Challenges Vitalik Buterin Over AI Risks to Lattice Cryptography
Cardano founder Charles Hoskinson has challenged Ethereum co-founder Vitalik Buterin over the security of lattice-based cryptography, escalating a debate about whether artificial intelligence could expose weaknesses in widely used encryption systems. Hoskinson argues that abandoning established post-quantum cryptographic research without a demonstrated attack risks undermining years of security engineering. The dispute comes as crypto developers reassess how AI-driven mathematical advances could affect digital assets and online communications.
Buterin warned on October 8 that AI-accelerated mathematics could threaten cryptographic systems and urged the industry to reduce exposure to potentially vulnerable methods. However, he explicitly advised against rushing to move funds into new wallets immediately. His warning concerns the possibility of faster mathematical breakthroughs, rather than confirmation that an existing cryptographic standard has been broken. <Cite refs={[“turn567911search9″,”turn567911search5”]}/>
Hoskinson disputes the argument that lattice-based encryption could contain an undiscovered mathematical shortcut comparable to breakthroughs in integer factorization. He says researchers have spent decades studying lattice attacks, including algorithms such as LLL, BKZ and lattice sieving. In his view, their known computational costs have already informed the security parameters selected for modern post-quantum systems.
He also challenges proposals to increase cryptographic key sizes dramatically as a precaution. Hoskinson argues that security parameters should be based on specific attack models, computational costs and measurable improvements in attack efficiency, rather than an arbitrary multiplier. His central point is that a credible threat assessment must identify a potential attack and quantify its consequences.
Hoskinson Defends Lattice-Based Post-Quantum Standards
A key part of the dispute concerns ML-KEM and ML-DSA, two lattice-based algorithms standardized by the US National Institute of Standards and Technology (NIST). ML-KEM establishes shared secret keys over public channels, while ML-DSA provides digital signatures. Both were selected through a multiyear evaluation process intended to prepare systems for future quantum-computing threats. <Cite refs={[“turn567911search1″,”turn567911search2″,”turn567911search4”]}/>
Hoskinson argues that the mathematical assumptions underpinning lattice cryptography have received extensive scrutiny. He points to research connecting attacks against Learning With Errors problems to difficult lattice problems, arguing that this gives the field a more developed basis for security analysis than simply assuming an algorithm remains safe because nobody has broken it.
He also questions the suggestion that hash-based cryptography is automatically safer because hashes are designed without exploitable mathematical structure. Hoskinson notes that cryptographic hash functions have suffered practical attacks, while emphasizing that the security of any specific construction depends on its design and the attack being considered. He argues that algebraic techniques used in some modern hash-based proof systems also deserve scrutiny.
The disagreement extends to proof systems and proposed changes to cryptographic infrastructure. Hoskinson contends that claims about AI discovering decades of mathematical breakthroughs within a short period do not constitute a usable threat model unless researchers identify a plausible method and estimate its cost. He favors continued testing and parameter adjustments over abandoning an entire cryptographic family in anticipation of an unspecified attack.
AI Security Debate Raises Questions for Crypto Infrastructure
Hoskinson’s argument does not mean lattice-based systems are immune to future discoveries. NIST’s post-quantum program continues to evaluate alternative algorithms, including the hash-based SLH-DSA signature standard and the HQC key-encapsulation mechanism selected for additional standardization work. Maintaining alternatives can reduce reliance on a single set of mathematical assumptions. <Cite refs={[“turn567911search1″,”turn567911search25”]}/>
The discussion also has implications for internet security. Lattice-based key establishment is relevant to protecting communications against future quantum attacks, including the risk that encrypted traffic collected today could be decrypted later. Hoskinson warns that unnecessary delays in deploying post-quantum protections could leave users exposed to existing vulnerabilities for longer.
There is also a distinction between theoretical concerns and demonstrated weaknesses. In July 2026, NIST reported that an AI model had helped discover a vulnerability in HAWK, a lattice-based signature candidate that was subsequently withdrawn from standardization consideration. NIST stated that the finding did not affect its finalized ML-KEM and ML-DSA standards. The episode illustrates why individual schemes must be assessed separately rather than treating every lattice construction as equivalent. <Cite refs={[“turn567911search0″,”turn567911search1”]}/>
For crypto developers, the immediate challenge is to evaluate new research without allowing either complacency or speculation to dictate security decisions. A credible attack would require independent verification, reproducible results and an assessment of which implementations and parameter sets are affected. Any necessary migration would then need to balance security, compatibility and operational costs.
The disagreement between Hoskinson and Buterin ultimately highlights a broader question for the industry: how should cryptographic systems respond to AI-driven advances in mathematics? Hoskinson argues for evidence-based changes grounded in established attack models, while Buterin urges greater caution about emerging risks. Until researchers demonstrate vulnerabilities in specific deployed standards, the debate remains a question of risk assessment and preparedness rather than proof that current lattice-based cryptography has been broken.
About The Author
