# Aztec Discloses Critical Soundness Flaw in Its Alpha v5 Proving System

Core contributors found on July 27 that an attacker could construct a proof the network would accept for a transaction it should reject, the second proving-system defect Aztec has disclosed this year.

- Published: 2026-08-19T05:47:33.890Z
- Canonical: https://polylog.news/crypto/2026-08-19/aztec-discloses-critical-soundness-flaw-in-its-alpha-v5-prov
- Publisher: Polylog (Crypto desk)
- Section: crypto
- Sources: [Aztec Network](https://aztec.network/blog/alpha-v5-proving-system-vulnerability), [Aztec Network (Alpha V5)](https://aztec.network/alpha-v5), [Miden](https://miden.xyz/blog/practical-privacy)

Aztec, the privacy-focused Ethereum layer-2 network, disclosed a critical vulnerability in the proving system behind its [Alpha v5 release](https://aztec.network/alpha-v5). Core contributors identified the defect on July 27 through internal auditing assisted by AI code analysis. The [disclosure](https://aztec.network/blog/alpha-v5-proving-system-vulnerability) describes a soundness failure: an attacker could build a proof that passes verification for a transaction the network's rules should reject.

Soundness bugs belong to a different risk class than smart-contract exploits. When a contract is exploited, the chain agrees on what happened, and the loss is bounded by the contract's balance. When a proving system accepts an invalid proof, the chain agrees on something false, and on a network where state is encrypted, the discrepancy is harder for outside observers to detect. Validator re-execution, the fallback most optimistic rollups rely on, does not apply to a system whose entire security argument is the proof itself.

This is the second such disclosure from Aztec in five months. In March the team disclosed a critical vulnerability affecting the Alpha v4 proving system and told v4 users to [withdraw funds before June 25](https://thedefiant.io/news/defi/aztec-v4-withdraw-june-25-v5-upgrade-security-vulnerability), when an on-chain governance vote forced public disclosure. Aztec says reviewers have not identified other high-severity or critical defects in v5 at this time, and its bug bounty on Cantina has drawn more than 234 researchers.

The pattern matches what the privacy sector itself argues. Miden, a competing privacy chain, writes that [most blockchains today are not private](https://miden.xyz/blog/practical-privacy) and that few users understand what that means in practice. The opposite risk is now just as visible: the systems that do deliver confidentiality carry a verification surface that only a small number of cryptographers can review, and repeated critical findings suggest that surface has not been fully tested.

## What this means

Anyone holding funds on a zero-knowledge network is trusting a proof system that no external re-execution can check, and Aztec has now found two critical defects in that system within five months. The exposure runs to depositors on privacy rollups and to the institutions being pitched confidential settlement, because a soundness break allows unbacked withdrawals rather than merely leaking data. Two outcomes are possible. Either proof-system diversity and formal verification become standard before institutional volume arrives, or a soundness bug is exploited on a network holding real balances, and the same confidentiality that hides account state also hides the theft.

## What to watch

- Whether Aztec publishes the full technical write-up and the fix timeline, since embargoed disclosure timed to upgrades leaves users deciding on partial information.
- Whether other zero-knowledge rollups adopt multi-prover setups or escape hatches, which would show teams treat proving-system bugs as expected rather than exceptional.
- How much AI-assisted auditing turns up next, given that both this finding and BitBox's firmware flaws came from internal AI review rather than external attackers.