# Aztec Discloses a Proving Flaw That Could Let Invalid Transactions Pass Verification

Core contributors found the defect through internal artificial-intelligence-assisted auditing, cannot rule out earlier exploitation, and have deferred the fix to the next network version.

- Published: 2026-09-24T05:46:16.721Z
- Canonical: https://polylog.news/crypto/2026-09-24/aztec-discloses-a-proving-flaw-that-could-let-invalid-transa
- Publisher: Polylog (Crypto desk)
- Section: crypto
- Sources: [Aztec Network](https://aztec.network/blog/alpha-v5-proving-system-vulnerability), [Aztec Network](https://aztec.network/alpha-v5), [Miden](https://miden.xyz/prifi-stack), [Miden](https://miden.xyz/blog/bread-is-privacy-baked-in</source)

Aztec, the privacy-focused layer-2 network on Ethereum, has [disclosed a critical vulnerability](https://aztec.network/blog/alpha-v5-proving-system-vulnerability) in the proving system of its [Alpha V5 release](https://aztec.network/alpha-v5). By the project's own description, an attacker could construct a proof that passes verification for a transaction the network is supposed to reject, and that transaction could then produce a state change outside the rules V5 is meant to enforce. Contributors say they cannot determine whether anyone exploited the flaw before they found it, and they are telling users to treat funds, applications and contract state on V5 as exposed until incident response finishes and operators complete the required network actions. The repair is planned for V6.

The category matters more than the individual bug. A smart-contract exploit abuses code that runs correctly. A soundness defect breaks the check that decides what counts as valid in the first place. On most rollups, validators re-execute transactions and can catch an invalid state root. On a network built to keep transaction contents private, that fallback is weaker by design, because the data needed to re-execute a transaction is the same data the system is built to hide.

Aztec says the finding came from internal auditing assisted by artificial-intelligence tooling, which creates a risk in both directions. The same capability that lets reviewers sweep proving code at scale is available to anyone looking for the next unsound constraint, and the first published result of that sweep was a critical flaw in a live network holding user funds.

The disclosure arrives while competitors make the same case to institutions. Miden has announced its [PriFi stack](https://miden.xyz/prifi-stack) for private digital finance and a self-custodial private wallet called [Bread](https://miden.xyz/blog/bread-is-privacy-baked-in), built on client-side execution. Privacy networks are now selling confidentiality as a prerequisite for serious on-chain business, which means the proof system is not a technical detail for their buyers. It is the entire security argument.

## What this means

Users and applications on Aztec's V5 network carry direct exposure, because a forged proof could move state without anyone detecting the move by re-executing hidden transactions. For the wider privacy sector the exposure runs through credibility: institutions evaluating confidential execution now have a documented case where the verifier, not the contract, was the weak point, and diligence will shift toward proof-system audits, fallback options and proof diversity rather than contract review alone.

## What to watch

- Whether Aztec operators complete the network actions the disclosure requires and whether any anomalous state transition is found in V5's history, which would separate a theoretical flaw from a realized loss.
- Whether other zero-knowledge rollups publish proving-system reviews or add fallback verifiers, a sign the industry now treats soundness as its own risk category.
- Whether Miden, Aztec and other privacy networks start disclosing who can see what under their compliance designs, since institutional buyers are asking for confidentiality and auditability at the same time.