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

Core contributors say an attacker could construct a proof that passes verification for a transaction the network should reject, and they cannot rule out that it was used before discovery.

- Published: 2026-08-27T05:48:48.749Z
- Canonical: https://polylog.news/crypto/2026-08-27/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](https://aztec.network/alpha-v5)

Aztec, the privacy-focused layer-2 network on Ethereum, disclosed a critical vulnerability in the proving system behind its [Alpha v5 release](https://aztec.network/alpha-v5). According to the project's own [disclosure](https://aztec.network/blog/alpha-v5-proving-system-vulnerability), an attacker may be able to construct a proof that passes verification for a transaction the network is supposed to reject, which would let the chain accept a state transition outside its own rules. Core contributors found the flaw through internal review, and they say they cannot determine whether anyone exploited it beforehand. Funds, applications and contract state on v5 should be treated as exposed until incident response finishes and operators complete the required network actions.

This is a different category of failure than a smart-contract bug. On a zero-knowledge rollup, the cryptographic proof itself certifies a transaction as valid, and there is no separate check behind it. If the verifier accepts an invalid proof, no amount of contract auditing would catch the error. The safeguard optimistic rollups rely on, independent re-execution of transactions by validators, does not exist in the same form here. That is why a soundness defect forces a network pause and a system upgrade rather than a patch to a single application.

This is not Aztec's first such disclosure. The project published a critical vulnerability notice for Alpha v4 earlier in its testing cycle and raised its bug bounty to $2 million. Both facts point to the same conclusion: this is alpha software, and the team treats pre-mainnet disclosure as the intended result of its review process, not as a failure.

This disclosure comes amid a broader wave of privacy-focused development. Aztec's own applications, including a live on-chain game and private decentralized finance (DeFi) integrations, run on the same proving stack that contains the flaw.

## What this means

Proof soundness has become the single point of failure for privacy rollups, and standard audit methods cannot detect it. A zero-knowledge layer-2 network cannot catch an invalid state transition by re-running transactions the way other systems can, so users, application developers and bridge operators depend entirely on the prover team's internal review and on whatever emergency mechanism the network provides. The parties exposed are depositors and applications on Aztec v5 during the response window, and, more broadly, every rollup that markets proof verification as a stronger guarantee than validator honesty. The mitigating factor is that this was caught during the alpha phase, before user funds reached mainnet scale.

## What to watch

- Whether Aztec publishes the technical root cause after the fix, since a specific circuit or verifier defect tells other teams whether their own stacks share it.
- Whether operators complete the required upgrade without any loss, which would support the case that staged alpha releases work as a safety process.
- Whether other zero-knowledge rollups announce independent verifier diversity or second proof systems, the practical defense against a single soundness bug.