Aztec Discloses a Proving-System Flaw That Let an Invalid Transaction Pass Verification

Aztec, the zero-knowledge privacy layer-2 network on Ethereum, has disclosed a critical vulnerability in the proving system of its Alpha v5 release. By the team's description, an attacker could construct a proof that passes verification for a transaction the network should reject. Core contributors identified the defect on 27 July 2026 through internal auditing assisted by artificial intelligence (AI) tooling, and the fix is scheduled for the next version rather than an emergency patch.

The distinction matters. A smart-contract bug lets an attacker misuse rules that the chain still enforces correctly. A soundness defect in the proving system breaks the enforcement itself, which means the chain can be made to accept state it would otherwise reject. Aztec has stated that this class of flaw is not caught by public re-execution by the validator committee, the check that ordinary optimistic and hybrid designs depend on.

This is the second such disclosure in six months from the same team. Aztec reported a critical flaw affecting the v4 proving system as a whole in March, warned that exploitation could allow theft of user funds, told v4 users to withdraw, and withheld technical details until the v5 upgrade shipped. Embargoed disclosure timed to an upgrade is now the standard practice for proving-system bugs across the zero-knowledge sector, because publishing the defect before the fix exists would give an attacker a working method.

Two readings compete, and both are available from the same facts. Aztec presents the finding as the intended output of an alpha network, meaning a live system carrying deliberately limited value while the proving stack is hardened. Critics of the rollup model see a pattern in which the security case for a chain rests on cryptography that only a small number of specialists can review, leaving users to trust a disclosure process instead of an independent check.