AdvertiseFee Deal Desk

Ethereum

Aztec CEO Can't Rule Out AI Finding New Bugs in Tierra V6 After V5 Flaw

Aztec CEO Zac Williamson says advanced AI models could still find bugs in Tierra V6, which launched Nov. 3 and patches a critical V5 proving-system flaw.

Be a creator
November 16, 2026, 04:23 AM UTC4 min read
AI SummaryAI
  • Aztec CEO Zac Williamson said advanced AI models could still find bugs in Tierra V6, launched Nov. 3.
  • Internal AI audits found a critical V5 proving-system flaw that could let forged proofs pass, disclosed in August.
  • Only Aztec's most security-critical subsystems are formally verified; the full protocol is not, per Williamson.
  • Major Aztec network upgrades run as complete hard forks, forcing users to withdraw and redeposit funds.
Flush

Tierra V6 Launched on Nov. 3

Aztec Foundation co-founder and CEO Zac Williamson has left open the possibility that newer AI models could surface undiscovered flaws in Aztec Tierra (Alpha V6), the latest version of the Aztec Network, which went live on Nov. 3. Speaking in an interview recorded on Nov. 9 and published on Nov. 15, he said he “can’t rule out” that more capable AI systems will find bugs in the release. The harness behind that confidence runs 24/7 on Aztec’s own infrastructure and turned up “some criticals” in the cryptographic libraries of the previous version, he said. In August, Aztec disclosed that AI-assisted internal auditing had caught one critical defect in the V5 proving system, one that could let a forged proof slip through verification, and said it could not establish whether anyone had exploited it. An Oct. 15 post he co-wrote stated that the AI audit harness had caught critical vulnerabilities, among them defects human reviews missed, and that V6 patches them. Tierra ships the fix, according to the launch post, which also introduced formal verification, the practice of mathematically proving that code behaves exactly as specified, covering the core cryptographic components that protect every private key. He framed the stakes in blunt terms: the cost of mounting cyber attacks is “falling to zero,” he said, and attackers can field “millions of AI bots.” In his telling, static defenses fall behind once offense scales cheaply. AI-assisted review has in any case become a fixture of modern blockchain security audits, a shift the industry has absorbed quickly. In September, Aztec Labs brought its zk.money payments wallet back to the network, giving the upgraded stack an early consumer product. The launch post also flagged faster private proving alongside the security fixes.

Upgrades Handled as Complete Hard Forks

Formal verification has not reached the whole protocol, however. Williamson said coverage stops at the “most security critical parts,” a boundary the team drew after the AI findings, and a step he said implies the bug the models caught was the last one in each affected subsystem. Model refusals come up rarely, he noted, because the audits hunt both cryptography and ordinary software defects, and Aztec runs uncensored open-source models “just to double-check.” Asked about a Zcash bug reported in June, Williamson said flaws in privacy systems can often create double-spend outcomes, in which the same funds are spent twice, with no visible trace on chain, a fault family that overlaps the replay attack. Zcash’s security advisory stated that turnstile accounting restricted how much value could exit the affected pool, and he called Zcash “best in class” at managing this kind of risk. Aztec copies the approach: major network upgrades are handled “effectively as complete hard forks,” meaning users must withdraw their funds from the old version of the protocol and deposit them again into the new one. For tokens that live on Ethereum, a hidden double-spend would surface during those withdrawals, he said, and it cannot extend into the new version of Aztec. ETHBEAT, the tracker that follows Ethereum layer-2 networks, records V6 as a clean break with none of V5’s state carried across, and gives it a Stage 2 rating, the highest tier it awards. Williamson also pointed out that Aztec operates without a security council of the kind other networks maintain, and named full formal verification “the long-term solution.” The same conversation took up how private protocols can prove that deposits are not linked to North Korea without exposing user identities, one of the risks Williamson was asked to weigh.

Containment Over Continuity

The picture that emerges is containment over continuity. Each hard-fork-style upgrade resets the ledger, which bounds the damage a latent double-spend can inflict on assets still parked on the old version, and the Zcash episode shows turnstile accounting held in practice. With no security council to appeal to, Aztec is betting that round-the-clock AI auditing, formal proofs over its most critical subsystems and enforced migration can do the work governance does elsewhere. Williamson’s own stated position closes the loop: formal verification of the entire protocol remains “the long-term solution,” and until Aztec gets there, every major upgrade will ask its users to move first.

Readers tracking the market in real time can follow live spot and futures prices on Bybit.

COINOTAG's editorial and research desk.

AI-Assisted

AI-generated, AI-reviewed, under COINOTAG editorial oversight.