Aztec Lost $4.3M to Exploits on Abandoned Code It Couldn’t Patch

Published by James Harris on June 18, 2026June 18, 2026

Two exploits hit Aztec’s abandoned smart contracts within four days, draining a combined roughly $4.3 million from infrastructure the team had formally walked away from years ago. The attacker didn’t need a zero-day or insider access. They needed patience and a read of old code.

The June 14 attack on the deprecated Aztec Connect bridge worked because the proof verification system and the on-chain settlement code disagreed on how many transactions in a batch were real. The proof system counted in groups of 32; the settlement layer trusted whatever number the batch declared. Fourteen crafted rollup submissions in a single transaction was enough to drain approximately 909 ETH, 270,513 DAI, 168 wstETH, and several Yearn vault tokens. The June 17 attack hit a separate contract entirely, a private rollup bridge that Aztec Labs described as “an immutable stage 2 rollup that was sunset in 2022.” Immutability, the property that makes zk-rollups credibly neutral, is exactly what prevented any patch. Aztec had renounced all administrative roles in April 2024 specifically to let remaining users exit without team interference, a decision that reads differently now.

Blockaid says its monitoring platform detected the attacker’s preparation activity roughly six minutes before the first drain executed. Six minutes, and no one with authority to act.

This is the underexplored liability of principled decentralization. When a team renounces admin keys and upgrade authority on-chain, they are making a permanent statement about trust minimization, but they are also permanently removing their own ability to respond to discovered vulnerabilities. The Aztec case is now a concrete data point in that tradeoff, and it arrives at a moment when DeFi’s June exploit losses had already crossed $43 million at the month’s midpoint, per DefiLlama. A third incident, a $2.1 million exploit on a legacy Thetanuts Finance vault on June 15, confirms this is a category of attack, not a coincidence. Protocols that migrated away from old contracts without draining them, or that renounced controls before all users had exited, are now the softest targets on-chain.

The practical implication for any protocol considering key renunciation is that the sequencing matters as much as the act. Renouncing before the contract is fully drained is not decentralization, it’s abandonment with a governance wrapper around it. As this pattern becomes more documented, expect security auditors and protocol reviewers to start treating “deprecated but funded” contracts as a distinct risk category requiring active wind-down, not just a deprecation notice.