# Aztec Discloses a Second Critical Proving-System Flaw, This Time on Its Live V5 Network

Core contributors found the defect through internal artificial-intelligence-assisted auditing and say an attacker could construct a proof that passes verification for a transaction the network should reject.

- Published: 2026-08-21T05:52:54.718Z
- Canonical: https://polylog.news/crypto/2026-08-21/aztec-discloses-a-second-critical-proving-system-flaw-this-t
- 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, has published a disclosure stating that its live Alpha V5 network contains a [critical vulnerability in the proving system](https://aztec.network/blog/alpha-v5-proving-system-vulnerability). Core contributors identified the defect through internal auditing assisted by artificial-intelligence tooling. The stated failure mode is soundness: an attacker may be able to build a proof that the verifier accepts for a transaction the network should have rejected. Aztec says users should 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.

This is the second consecutive Aztec release to carry a proving-system defect. The team disclosed a critical flaw in Alpha V4 earlier this year, told V4 users to withdraw before the V5 governance vote, and said at the time that the bug touched the proving system as a whole and could lead to theft of user funds. [Alpha V5](https://aztec.network/alpha-v5) was the version meant to close those findings. Instead, the fix cycle produced a fresh soundness bug in the same layer.

A soundness bug is a different risk class from a smart-contract exploit. On an optimistic rollup, honest nodes re-execute transactions and can dispute an invalid state root. On a validity rollup, proof verification is the only check on whether a state transition is valid before it becomes final, and no re-execution fallback catches a verifier that accepts a bad proof. That is why disclosure gets embargoed until an upgrade is ready, and why the interval between discovery and remediation leaves users unable to independently assess the risk during that interval.

The timing sits against a broader push by privacy networks to present confidential execution as a prerequisite for institutional on-chain activity. Miden, a competing privacy chain, argues in its own material that [most blockchains today are not private](https://miden.xyz/blog/practical-privacy) and that the general public does not yet recognize what public state actually exposes. That pitch depends on the underlying proof machinery being trustworthy. Every soundness disclosure raises the evidentiary burden these networks must clear before a bank or an asset manager routes real balances through them.

## What this means

Validity rollups concentrate risk in one component, the proof verifier, and Aztec has now found defects there twice in a row. The exposed parties are holders of assets and applications on the live V5 network, who cannot verify the fix themselves and cannot rely on node re-execution to catch a forged proof. For the wider privacy-chain sector, the channel is credibility: institutions evaluating confidential settlement will price a longer audit and observation period into any deployment, which slows the revenue timeline for privacy layer-2 teams and favors chains that can point to independent proof-system diversity or a working escape hatch.

## What to watch

- Whether Aztec publishes a full technical post-mortem naming the specific circuit or verifier component at fault, which would tell researchers whether the V4 and V5 bugs share a root cause or are independent.
- Whether other validity rollups respond by adding a second, independently implemented proof system or a user-triggered exit path, a sign the sector now treats single-prover designs as unacceptable.
- How much value stays on Aztec through the remediation window, which measures whether users actually act on protocol-level warnings or ignore them.