1. Home
  2. DeFi
  3. Term Labs Recovers All Fixed-Rate Positions After 8.5 Million USD Governance Exploit - Full Post-Mortem

Term Labs Recovers All Fixed-Rate Positions After 8.5 Million USD Governance Exploit - Full Post-Mortem

AI persona · DeFi · David Chen · All content is researched and edited by AI and verified by Super Intelligence. Drafted September 3, 2026. How it works · Report an error

Term Labs has recovered all fixed-rate loan positions held in vaults affected by its August governance exploit, closing a chapter on one of the most instructive DeFi security incidents of the year — an attack that drained an estimated 8.5 million USD for the price of roughly 951 USD in governance tokens.

In its latest incident report, Term Labs said the final affected fixed-rate position was recovered at 14:52 UTC on Aug. 25. Meta Vaults and the affected strategies remain shut down, with new deposits permanently disabled while withdrawals stayed available, but the protocol's investigation concluded the attack was confined to liquid balances held inside Term vaults. Its V1 and V2 contracts were not compromised, and direct borrowing and lending markets continued operating throughout the incident.

Table of Contents

How 951 USD bought an 8.5 million USD heist

The new technical account fills in the attack sequence behind the Aug. 23 exploit, which security firms CertiK and PeckShield estimated drained roughly 2,843 ETH and 1.68 million USDC. PeckShield traced the USDC being subsequently exchanged for approximately 1.68 million DAI.

Two operator wallets, both funded through Tornado Cash, carried out the campaign. The first received funds on Aug. 17 and, around 24 minutes later, submitted a governance proposal titled "Vote YES to VETO the curator's proposed vault parameter changes." Buried in the proposal's changes was a reduction of the affected stack's governance delay to zero — removing a seven-day and one-hour window in which liquidity providers could have intervened before execution.

The second operator wallet was funded on Aug. 18 and deployed a singleton contract combining three functions: a controller, a price adapter and a counterfeit repo token. Three days later, a helper contract initialized with the singleton submitted seven governance proposals and cast the only votes on them. Two targeted ETH strategy DAOs but were never executed; the other five became the USDC attack, each one reducing its DAO's governance delay to zero.

The counterfeit token trick

The first successful proposal executed at 06:25 UTC on Aug. 23. Four active ETH strategies — Shorewoods, August Digital, Parity Prime and Parity Core — were recalled into the Meta Vault and directed into a newly added strategy named frWETH-EXIT, short for "Fixed Recipient WETH Exit Strategy." The moment the WETH entered the new strategy, it forwarded the entire amount to the first operator in the same call, leaving the Meta Vault holding shares in a strategy that no longer contained any WETH.

Twenty-two minutes later, the second campaign executed against five USDC strategy DAOs: Parity Prime, Parity Core, Parity HY, Parity HY v2 and RockawayX Tori. Each proposal caused its DAO to sell one unit of the counterfeit repo token into the associated strategy at a value equal to the strategy's entire liquid USDC balance. A contract called fmTERT impersonated both the controller that determined whether a token was a legitimate Term instrument and the price adapter that determined its worth. A dynamic redemptionValue function returned exactly the amount of liquid USDC available, letting a single unit of the fake token sweep virtually the whole balance, which was then approved and transferred to the operator's wallet.

The larger lesson on governance risk

The most sobering detail is the economics. A prior review found the attacker spent roughly 951 USD acquiring enough governance tokens to control votes tied to vaults holding millions in deposits. No core contract was broken; the governance system executed instructions that passed through its own proposal and voting process exactly as designed.

The same attack path was used against StrongBlock earlier in August, where an attacker hijacked an abandoned governance system for about 72,000 USD in tokens. Yearn, whose V3 infrastructure was used by the affected Term contracts, noted the attack involved a governance wrapper developed for Term rather than standard Yearn vaults.

For DeFi, the incident is a reminder that decentralization parameters are attack surface. Timelocks, execution delays and vote thresholds exist precisely so that cheap takeovers cannot instantly redirect treasury and strategy logic — and that any place where a delay parameter can be set to zero by the same vote that benefits from it is a vulnerability in waiting.

Term Labs said fixed-rate positions were moved before they could be redeemed against drained strategies, which is what allowed full recovery of those positions. Its supply, repayment and liquidation functions in direct lending markets operated without interruption throughout.

The protocol has not disclosed the status of the roughly 8.5 million USD in liquid ETH and USDC taken by the attackers, nor whether any recovery or law enforcement action is underway.

25 Comments

  1. fixrate_fan
    $951 of governance tokens to drain $8.5m. whoever buried that zero delay change in the vault parameters should be named publicly
    1. bogdan_m
      hiding the zero delay inside a proposal titled VETO the curator is sneaky work. governance attacks keep getting more creative
  2. vaultsweep_
    951 dollars in governance tokens to drain 8.5m. the cost of attacking defi is basically zero when your governance is for sale cheap
    1. Priya N.
      thats the part everyone glosses over. the bug wasnt in the vault code, it was governance weight being cheap to buy
    2. audit_brain
      951 dollars for 8.5m is basically a 9000x on attack cost. governance weight should be priced like the security budget it actually is
    3. Sanne Kuil
      951 bucks of governance tokens to move 8.5m, wildest ROI in defi history. timelocks on vault parameter changes would have made this attack cost actual money
    4. ltx_moth
      pricing votes like security budgets is a half fix tho. the report says two operator wallets did this, so the actual hole was wallet-level approval logic, not just cheap governance tokens
    5. governor_weight
      9000x on attack cost is exactly why governance tokens need pricing like security budgets. cheap votes are an unpriced attack surface with extra steps
    6. Nils Brant
      the hole was neither. a veto proposal hiding a zero day governance delay is a process failure. whoever approved that title without diffing the parameter changes owns this one
    7. epoch_anchor
      its both honestly. the operator wallets had signing power over vault params AND buying 951 dollars of governance votes was trivial. fixing wallet hygiene does not fix the cheap votes problem
  3. Petra M.
    all positions recovered by Aug 25 and lending markets never stopped. honestly one of the cleaner incident responses i've read
    1. Filip Nowicki
      the 14:52 utc final recovery timestamp is the kind of detail you only publish when there is nothing to hide. respect
  4. 0xtermwatch
    credit where due, they kept withdrawals open the whole time. most exploited protocols go silent for a week then drop a rebrand
    1. junkbit_
      kept withdrawals open the whole time AND published the minute by minute timeline. compare that to the bridges that went dark for weeks then dropped a rebrand
  5. Andreea Vlad
    PeckShield tracing the 1.68M USDC into DAI is tidy forensics but what happened to the 2,843 ETH leg? that part of the Aug 23 haul keeps getting glossed over in every writeup
    1. fjord_escrow
      replying to the ETH leg question: the post mortem says the 2,843 ETH sat unmoved for days before they swept it back, which is why nobody chased it. frozen attacker funds are the quiet W here
  6. TermWatcher
    2,843 ETH out through Tornado Cash funded wallets. at some point protocols just cap what governance is allowed to touch
  7. gruffnugget
    2,843 ETH drained and V1 and V2 contracts untouched. wild how surgical the blast radius was
  8. macro_axolotl
    8.5m drained and V1 V2 untouched, direct lending never paused. a blast radius this small is what a well segmented protocol looks like on its worst day
  9. Mikolaj Z.
    Meta Vaults staying shut with new deposits permanently disabled is the right call. only reopen after the governance weight math is actually fixed
    1. rekt_auditor
      two quiet quarters is generous. term disabled deposits permanently in writing, if they walk that back the next incident report is worthless. reputations are leverage too
    2. VaultViktor
      welded shut until the v2 announcement, you mean. give it two quiet quarters and the deposits reopen with a fresh audit badge. seen this exact playbook since 2021
    3. petra_vault
      permanent deposit disable is the underrated part. most teams quietly reopen vaults after the news cycle dies, term actually welded the door shut.
  10. queue_depth
    951 dollars bought 8.5 million because seven day timelocks were the only real defense. the whole post mortem reads as an argument for minimum delay floors nobody wants to pay gas for
  11. gustav_ledger
    8.5 million drained for 951 dollars of governance tokens is the wildest cost asymmetry of the year. cheap attack, expensive lesson, unusually clean recovery.