# Aztec Discloses a Critical Soundness Flaw in Its Alpha V5 Proving System

Core contributors found the defect on July 27 through internal auditing assisted by artificial intelligence (AI), and say an attacker could construct a proof the network accepts for a transaction it should reject.

- Published: 2026-08-30T05:48:31.718Z
- Canonical: https://polylog.news/crypto/2026-08-30/aztec-discloses-a-critical-soundness-flaw-in-its-alpha-v5-pr
- Publisher: Polylog (Crypto desk)
- Section: crypto
- Sources: [Aztec Network (vulnerability disclosure)](https://aztec.network/blog/alpha-v5-proving-system-vulnerability), [Aztec Network (Alpha V5)](https://aztec.network/alpha-v5), [Aztec Network (applications live on Aztec)](https://aztec.network/blog/dark-forest-aztec-game-goes-live), [The Defiant](https://thedefiant.io/news/defi/aztec-v4-withdraw-june-25-v5-upgrade-security-vulnerability)

Aztec, the privacy-focused Ethereum layer-2 network, has [disclosed a critical vulnerability in the proving system behind its Alpha V5 release](https://aztec.network/blog/alpha-v5-proving-system-vulnerability). Core contributors identified the defect on July 27 during internal auditing assisted by AI tools. According to the disclosure, an attacker may be able to build a zero-knowledge proof that the network's verifier accepts for a transaction the rules should reject. Contributors advised users to treat funds, applications and contract state on [Alpha V5](https://aztec.network/alpha-v5) as exposed to a protocol-level failure until incident response finishes and network operators complete the required actions.

This is a soundness failure, not an ordinary smart-contract bug. A rollup that proves its state transitions replaces the work every node would otherwise redo. When the proof system itself accepts an invalid statement, the fallback most networks rely on, independent re-execution by validators, does not exist to catch it. That is the reason this class of defect keeps arriving through embargoed disclosure timed to an upgrade rather than through an on-chain incident.

This is not Aztec's first disclosure of this kind. In March, the team disclosed a critical flaw affecting the entire Alpha V4 proving system and told V4 users to withdraw funds before June 25. That date mattered because it was when the V5 governance proposal was set to move on-chain, a step that would force public disclosure of the older bugs under the project's transparency commitments. Two consecutive alpha releases have now carried a critical proving-system defect, and both were found by the people building the system rather than by an attacker.

The network is not idle infrastructure. Applications have been shipping on it, including [a live on-chain game built around hidden state](https://aztec.network/blog/dark-forest-aztec-game-goes-live), the kind of use that only works when execution is private. Aztec's pitch, and the pitch of every confidential-execution project competing with it, rests on users believing the cryptography holds. Each disclosure of this kind moves the conversation about privacy chains from what they can hide to whether their verifiers can be fooled.

## What this means

Zero-knowledge proving systems are now a distinct risk layer sitting under both privacy chains and general-purpose rollups, and it is a layer that audits scoped to application contracts do not reach. Anyone holding assets on a proof-secured network is exposed to a single verifier implementation, and the practical defense is a withdrawal mechanism and an operator able to halt the chain, not a bug bounty. The two outcomes worth separating: either teams normalize AI-assisted internal review that finds these defects before attackers do, in which case disclosure waves like this one keep arriving without losses, or one such flaw is found first by an attacker, in which case a proof-secured network loses funds with no re-execution fallback to stop it.

## What to watch

- Whether Aztec publishes the technical detail of the defect after network operators upgrade, which will show whether the flaw was in the circuit logic or in the verifier, and whether other proof systems share the pattern.
- Whether other zero-knowledge rollups run comparable AI-assisted audits of their own provers and disclose results, which would tell you if this is an Aztec-specific engineering gap or a sector-wide backlog.
- Whether users and applications on privacy networks demand proof-system diversity or exit mechanisms as a condition of deploying, which would raise the engineering cost of launching a new confidential chain.