# Aztec Says a Flaw in Its Alpha V5 Proving System Can Validate Transactions the Network Should Reject

Core contributors found the defect on 27 July through internal AI-assisted auditing, cannot rule out prior exploitation, and have scheduled the fix for the next network version.

- Published: 2026-09-13T05:57:01.509Z
- Canonical: https://polylog.news/crypto/2026-09-13/aztec-says-a-flaw-in-its-alpha-v5-proving-system-can-validat
- Publisher: Polylog (Crypto desk)
- Section: crypto
- Sources: [Aztec Network (disclosure)](https://aztec.network/blog/alpha-v5-proving-system-vulnerability), [Aztec Network (Alpha V5)](https://aztec.network/alpha-v5), [Aztec Network (gas design)](https://aztec.network/blog/how-gas-works-on-aztec)

Aztec Network, an Ethereum layer-2 built for private smart contracts, has [disclosed a critical vulnerability in the proving system](https://aztec.network/blog/alpha-v5-proving-system-vulnerability) that secures its live [Alpha V5 network](https://aztec.network/alpha-v5). Core contributors identified the defect on 27 July 2026 during internal auditing assisted by artificial intelligence tools. According to the disclosure, an attacker may be able to construct a proof that passes verification for a transaction the network is supposed to reject, which would allow a state transition outside the rules the protocol intends to enforce.

Aztec told users to treat funds, applications and contract state on V5 as exposed to a protocol-level failure until contributors finish incident response and operators complete the required network actions. Contributors said they cannot determine whether anyone exploited the flaw before they found it. Reviewers have not identified other critical or high-severity findings in V5, and the correction is planned for V6, the next version of the network.

This is the second proving-system failure Aztec has disclosed this year. It [warned V4 users to withdraw](https://thedefiant.io/news/defi/aztec-v4-withdraw-june-25-v5-upgrade-security-vulnerability) before publicly disclosing an earlier critical flaw, and it has since raised its bug bounty to two million dollars.

The class of bug matters more than the individual product. A soundness defect in a zero-knowledge proving system is not a smart-contract bug. The fallback that most optimistic rollups rely on, honest parties re-executing transactions and challenging bad state, does not exist in the same form for a validity-proof chain, because the chain accepts whatever the verifier accepts. That places unusual weight on a small number of cryptographic implementations, and on the [fee and operator design](https://aztec.network/blog/how-gas-works-on-aztec) that decides who can push a proof on-chain in the first place. Aztec's own account is also a data point in a separate argument: the flaw was surfaced by machine-assisted review rather than by an attacker.

## What this means

Validity-proof chains concentrate their entire security budget in one verifier. When that verifier can be fooled, every application and every balance on the network inherits the defect at once, and users have no re-execution fallback to catch it. The exposed parties are depositors on Alpha V5 and the teams building on it, plus the wider set of zero-knowledge rollups whose users now have to price proving-system risk as a separate line item from contract audits. Two outcomes are possible from here: Aztec ships V6 with no evidence of exploitation and the episode reinforces the case for AI-assisted auditing, or on-chain analysis later shows an anomalous state transition, which would make prover diversity and escape hatches a hard requirement rather than a research topic.

## What to watch

- Whether Aztec publishes on-chain evidence that no invalid state transition occurred on V5, which would tell users whether the disclosure was a near miss or an incident.
- Whether other zero-knowledge rollups respond with second verifier implementations or delayed upgrade windows, a sign the industry now treats prover soundness as a standing risk rather than a one-off.
- How much value stays on V5 in the days after the disclosure, which shows whether users actually act on protocol-level warnings.