Solana co-founder Anatoly Yakovenko has warned that artificial intelligence could expose weaknesses in post-quantum signature schemes, posing a risk that may arrive before quantum computers themselves. In a recent discussion, he said the biggest threat to the cryptography designed to survive quantum attacks might not be quantum hardware, but the machine learning models trained to crack those defenses.
The AI Risk to Post-Quantum Cryptography
Yakovenko didn't mince words. “I think the biggest risk is that pqc signature schemes will get broken by ai,” he stated. While quantum computing is still “years away” according to Solana's official messaging, AI is already here and improving fast. That shifts the timeline — and the threat model — for blockchain networks working to future-proof their security.
Post-quantum schemes rely on mathematical problems that are hard for classical and quantum computers alike. But AI, with its ability to find patterns and shortcuts, might find exploits that conventional attacks miss. Yakovenko's concern is that researchers won't see the flaw until it's too late.
Solana’s Quantum-Readiness Push
Solana has been preparing for a post-quantum world for some time. The network's research converged around the Falcon digital signature scheme, independently identified by both the Anza and Firedancer teams. The plan includes integrating Falcon signatures, migration planning for validators and wallets, and building resilience at the wallet level so users can upgrade without a network-wide overhaul.
The migration work is described as “well-researched, understood, and ready to deploy,” even if quantum computers themselves remain distant. But Yakovenko's warning suggests the team may need to accelerate parts of that timeline if AI poses a nearer-term danger.
Progress on Falcon Implementation
Developer Dean Little highlighted recent progress on the Solana Falcon implementation. Version 0.1.2 requires roughly 173,000 to 183,000 compute units (CUs) to verify a signature. That's a concrete step toward making Falcon practical on Solana, though the cost is still higher than the current Ed25519 signatures used for most transactions.
Lowering the verification cost is key to wide adoption. Validators and applications need assurance that switching to post-quantum signatures won't clog the network or raise fees dramatically.
Yakovenko’s Call for Deeper Native Support
Yakovenko also suggested a deeper architectural change: a syscall that lifts the PDA is_signer check to the transaction processor and charges fees to valid signers at the end of the block. That would allow the protocol to handle post-quantum signatures more natively, rather than forcing each wallet or dApp to manage the complexity alone.
Such a change would require a Solana improvement proposal and likely a network upgrade. It's not a trivial lift, but Yakovenko sees it as necessary for a smooth transition.
Defense-in-Depth Strategy
When asked about formal verification of the post-quantum code, Yakovenko was measured. “If we know exactly what to verify. I’d still like 2/3 different signature schemes,” he said. Rather than betting everything on Falcon, he prefers using two or three distinct signature schemes for defense-in-depth. That way, if one scheme gets broken — whether by quantum computers or AI — the others still protect the network.
The strategy echoes best practices in cryptography: never rely on a single algorithm. Solana's roadmap may include multiple post-quantum signatures alongside the existing Ed25519, giving users and validators options.
At press time, SOL traded at $84.03. The timeline for deploying Falcon on mainnet remains unclear, but Yakovenko's comments add a new urgency to the conversation: prepare for AI, not just quantum.
