Aave wasn’t hacked. Why did its users face restrictions?
A Binance-linked attacker borrowed from Aave before emergency controls took effect. The response stopped new deposits as lenders withdrew, pushed up borrowing costs and reopened a debate over how much disruption emergency safeguards should impose on other users.
Summary
- A wallet linked to the Binance attacker deposited 60 million USDT into Aave and borrowed 19,762 WETH, worth approximately $44.6 million, before emergency restrictions took effect.
- Aave’s guardians froze new deposits and borrowing across 175 V3 reserves and 117 V4 reserves, while withdrawals, repayments and liquidations remained available.
- Aave announced on October 30 that all V3 and V4 markets had reopened, stating that its contracts were unaffected and existing positions remained fully collateralised.
- Risk stewards removed approximately $3.83 billion in unused borrowing capacity to limit the size of new positions that could be opened when markets resumed.
- Nine assets, including USDT and USDC, remained unable to provide new borrowing power as collateral, although users could still supply, withdraw, borrow and repay them under the remaining market rules.
Aave has reopened its V3 and V4 lending markets after emergency restrictions following the Binance hack, while keeping 9 assets from providing new borrowing power as collateral.
Aave said in its October 30 update that its contracts were unaffected and positions remained fully collateralised throughout the incident. The restrictions followed a Binance-linked attacker’s use of the protocol to borrow wrapped Ether against stolen USDT.
The attacker had already taken out the loan before guardians froze new activity. Customers with no connection to the theft then faced restrictions on adding collateral while their debts continued accruing interest.
Lenders could still withdraw available funds, and borrowers could repay. New deposits were blocked, however, so the higher interest rates that normally attract money back into a lending pool could not bring in fresh supply.
Risk provider LlamaRisk responded with temporary rate changes and tighter borrowing limits before the reopening. Its October 30 account describes a response that had to contain suspicious funds and manage the pressure its own restrictions placed on existing customers.
The dispute has drawn in rival lending protocol Morpho, whose co-founder defended isolated markets, and Aave supporters who argue that emergency intervention provides protections an immutable market cannot offer.
The attacker borrowed before Aave froze new activity
The loan began with a deposit of 60 million USDT into Aave V3’s Ethereum Core market between 20:58 and 21:01 UTC on October 29, according to LlamaRisk’s incident account.
A wallet in the Binance exploiter’s cluster then borrowed 19,762 WETH, worth approximately $44.6 million at the time. The first freeze transaction followed at 21:42 UTC.
The transaction used Aave’s collateralised lending system. An asset deposited by the borrower supported a loan in another asset, leaving the collateral in the protocol as the borrowed WETH moved into the borrower’s control.
LlamaRisk put the position at $59.9 million of USDT collateral against $43.8 million of debt in its initial assessment. It said the loan remained overcollateralised, was subject to normal liquidation and did not put suppliers at risk.
Aave’s intervention was therefore precautionary. Its contracts had not been breached, and the attacker had not obtained an unsecured loan through a failure in the lending code, according to the platform’s account.
The stolen collateral originated in the Binance hot wallet breach, which the exchange estimated at roughly $6.3 billion. Binance said it would cover the losses with company resources and its user-protection fund.
An exchange’s reimbursement promise does not reverse a loan recorded on a separate protocol. The USDT deposit and WETH debt remained an Aave position subject to the lending market’s rules.
As crypto.news previously reported, the emergency freeze drew criticism after Emilio, posting on X, explained that restricting borrowing alone would still leave attackers able to deposit assets. He defended stopping deposits as a way to prevent those funds from entering the markets and complicating recovery.
The response extended across assets because the attacker could exchange stolen tokens for other listed collateral. LlamaRisk said the team chose a reversible guardian freeze over an initial cap change that could have taken several days to restore through governance.
A reserve freeze left debts and liquidations running
The restrictions stopped new supply and borrowing across the affected reserves. They did not cancel existing debt or suspend the liquidation process.
Aave’s Pool Configurator documentation treats freezing a reserve and pausing a reserve as separate controls. The incident response used freezes, and LlamaRisk explicitly said no market had been paused.
The distinction affected how customers could manage existing positions. A borrower could reduce debt by repaying, but could not supply more tokens to a frozen reserve to reinforce the account’s collateral.
Under Aave’s normal rules, either action can improve a loan’s health factor. Its liquidation guide describes that measure as collateral value, adjusted for the applicable liquidation thresholds, divided by the value of outstanding borrowing.
A health factor below 1 makes a position eligible for liquidation. The calculation can deteriorate when collateral falls in value, the borrowed asset rises or interest adds to the debt.
During the freeze, repayments remained the available way to improve that measure within a frozen reserve. A customer who ordinarily would have transferred extra collateral had to obtain the asset needed to repay instead.
Repayment can be partial, according to Aave’s repayment documentation. It reduces the outstanding loan without requiring the borrower to close the entire position.
Suppliers retained a withdrawal route, but the amount available depended on unborrowed liquidity. Aave’s withdrawal guide says customers can withdraw supplied assets only to the extent the pool has funds available.
Collateral supporting an open loan has another constraint. A withdrawal must leave the account with enough backing for its debt, so a functioning withdrawal transaction does not mean every customer can remove their entire balance immediately.
LlamaRisk said permissioned collateral kept the Horizon real-world asset market outside the freeze. The response covered 175 reserves on V3 and 117 on V4, with guardian transactions executed between 21:42 and 23:04 UTC.
Interest rates rose while fresh deposits were blocked
The rate pressure appeared in pools that still had collateral backing their loans. Lenders withdrawing available funds left a larger share of the remaining supply borrowed, pushing utilisation higher.
Aave’s borrowing guide says variable interest rates respond to utilisation and governance parameters. The rate shown when a borrower opens a position can change as lending and repayment activity alters conditions in the pool.
Higher rates ordinarily make lending more attractive and borrowing less attractive. During a deposit freeze, the first response was unavailable because suppliers could not add new funds to the affected reserves.
LlamaRisk made that point in its October 30 interest-rate proposal. It said elevated rates were adding to borrower debt and weakening health factors without attracting deposits.
Its snapshot, taken at 00:26 UTC, showed 97.8% utilisation in V3 Core’s USDC reserve and a 10.8% annualised borrowing rate. USDT0 on Plasma was at 100% utilisation with a 14.5% rate, while USDC on Monad stood at 98.1% utilisation and 20.2%.
The provider proposed reducing the part of the interest-rate curve called Slope2, which governs the steeper pricing above the pool’s target utilisation. For the selected USDC reserves, its table showed rates falling to 8% on Core and 8.8% on Monad at the recorded utilisation levels.
An annualised rate is a pace of accrual, not an immediate charge of that percentage of a customer’s debt. A temporary jump in the displayed APR therefore cannot be reported as an equivalent percentage loss during the hours of the freeze.
LlamaRisk said in its 13:49 UTC update that risk stewards had executed temporary rate adjustments. It said the settings would be reviewed as markets returned to normal.
The intervention did not amount to a refund of interest already accrued. The provider described changing the rate settings, without announcing a programme to reimburse every borrower for the period of restricted activity.
Borrowing limits addressed capacity available at reopening
The next problem concerned the unused borrowing capacity that would become available when the freezes ended. A reserve could reopen with a large gap between its existing debt and the maximum permitted borrowing.
In its separate borrowing-cap proposal, LlamaRisk said an attacker with remaining assets could use that gap to build a large position in the first blocks after reopening.
The proposed limits took the higher of current borrowing and borrowing immediately before the exploit, then added 2%. The provider said the changes removed approximately $3.7 billion of unused V3 capacity and $130 million on V4.
The V3 Core table reduced the WETH cap from 2.4 million to 1.79 million tokens, USDT from 3.09 billion to 2.56 billion and USDC from 2.7 billion to 2.27 billion.
Subtracting the published limits gives reductions of 610,000 WETH, 530 million USDT and 430 million USDC. Dividing each reduction by its previous cap gives cuts of approximately 25.4%, 17.2% and 15.9%, respectively.
The calculations describe borrowing ceilings, not withdrawals, repayments or losses. The provider said existing debt remained below the new limits and each reserve could return to its pre-exploit borrowing size.
LlamaRisk’s implementation update said the cap batch had been executed, while GHO changes on V3 Core and Prime had been queued with the GHO Stewards.
Reopening did not restore every asset’s collateral permissions
Aave’s October 30 reopening statement said all V3 and V4 markets were unfrozen. The announcement retained restrictions on USDT, USDC, USDT0, POL, WAVAX, sAVAX, XAUt, PAXG and OP.
The restrictions concern using the assets to support new borrowing. They are not a ban on holding or transferring the tokens, and they do not mean the assets have a market price of zero.
Aave’s reserve documentation defines loan-to-value as the borrowing power assigned to collateral. Liquidation thresholds serve a separate function in determining when an existing position can be liquidated.
LlamaRisk’s rollback update distinguished how the retained restrictions work in the protocol’s 2 versions. V3 uses an LTV of 0, while V4 assigns a collateral factor of 0 for new borrowing against the assets.
On V3, existing positions retain their liquidation thresholds, so the LTV change alone does not alter their health factors. The restriction applies to E-mode as well, which ordinarily offers tailored borrowing parameters for selected asset categories.
On V4, existing positions remain on their current collateral configuration until a new borrow, collateral withdrawal or collateral toggle moves them to the new configuration. A customer changing an existing position can therefore encounter a restriction that was not applied to it before that action.
The provider said the keep-list covers assets held across the exploiter’s address cluster, including bridged and wrapped versions. Supplying, withdrawing, borrowing and repaying those assets can continue, subject to the normal market rules and available capacity.
The restrictions do not single out a customer based on whether their own tokens came from Binance. They apply to the specified collateral across markets where it is listed.
LlamaRisk said it would assess the conditions for restoring borrowing power and publish them in the incident thread. Its update did not give a fixed date for removing the remaining collateral controls.
Morpho’s defence focuses on containing exposure by market
Not everyone in DeFi agrees that the ability to intervene across markets offers the best protection for users. Morpho co-founder Paul Frambot defended a design in which participants choose their exposure through isolated lending markets.
In his October 30 comments, Frambot said trouble in one market does not automatically spread to others. He acknowledged that individual markets could still experience liquidity constraints and volatile interest rates.
Morpho’s market documentation describes a variable-rate market as a pool pairing 1 collateral asset with 1 loan asset. Each market has its own liquidation loan-to-value ratio, oracle and interest-rate model.
The parameters are selected at creation and remain immutable. Lenders choose the market in which to supply the loan asset, while borrowers deposit the specified collateral and borrow within its limits.
A lender in another isolated market does not share the same loan simply because both markets use Morpho. Someone allocated to the affected market, however, still has exposure to its collateral and borrowers.
Marc Zeller challenged Frambot’s argument, saying Morpho Blue lacks the freeze powers and ability to intervene in the interest-rate curve available in Aave’s approach. He argued that isolation does not protect the lenders inside a market that fails.
Morpho’s interest-rate documentation confirms that a market’s interest-rate model address cannot be changed by governance or its creator. Its approved AdaptiveCurveIRM adjusts automatically as supply and borrowing conditions change.
The choice of model therefore constrains what can be altered during a disruption. Aave’s stewards changed rate parameters during this incident; Morpho Blue’s design relies on the behaviour of the model selected when each market was created.
Zeller pointed to Aave’s Umbrella protection system as another difference. According to Aave’s Umbrella guide, participants stake assets in return for rewards while accepting that the stake can be used to cover protocol deficits.
The protection is assigned to particular assets and networks. The documentation describes automated use of the corresponding stake when a qualifying deficit occurs, rather than an unlimited promise covering every interruption to lending.
Neither design makes liquidity constraints disappear. Morpho’s withdrawal rules require available liquidity in the individual market, just as Aave suppliers depend on funds available in the reserve from which they withdraw.
Frambot’s case rests on choosing and containing exposure before a problem occurs. Zeller’s defence rests on the ability to intervene once a problem reaches the lending system, and on resources available if collateral no longer covers debt.
April’s recovery shows why a loan and recovered funds differ
Aave had already dealt with the consequences of an attack on another platform earlier in 2026. The Kelp DAO incident involved rsETH used as collateral after a bridge exploit.
Glassnode’s April 27 liquidity study found that available liquidity on Aave V3 Ethereum Core fell from $9.77 billion to $5.75 billion within 29 hours. It said the contracts, oracles and liquidation engine operated as specified during the episode.
The research put available WETH liquidity at $1.5 million after a fall from $689 million in 2 hours. A functioning protocol had continued processing transactions as liquidity available to its users contracted.
The collateral problem in April differed from the initial assessment of the Binance-linked loan. The Kelp recovery concerned impaired backing for rsETH, while LlamaRisk described the October USDT position as overcollateralised in its initial account.
Recovery required more than switching lending back on. Aave Labs reported the final attacker-position liquidations in May, with recovered collateral transferred to the DeFi United recovery group’s guardian wallet.
The recovery operation did not activate Umbrella, according to Aave’s statement at the time. Support from other protocols and the treatment of frozen assets formed separate parts of the effort to restore rsETH backing.
A disputed Ether transfer involved 30,765 ETH worth roughly $71 million at the time. A Manhattan court modified a restraining notice to allow the governance process to proceed while preserving creditors’ legal claims.
The funds had been frozen by Arbitrum’s Security Council after the exploit. A governance decision about their transfer and a court dispute over claims to them followed the technical intervention.
Aave restored WETH borrowing across affected markets in May as the recovery progressed. The reinstatement of borrowing permissions and the handling of recovered assets followed different procedures and timetables.
The October risk measures have not been presented as an announcement of bad debt equivalent to April’s. LlamaRisk’s account identifies a collateralised loan and preventive controls, without establishing that the same recovery process will be required.
Automated limits have returned to the governance discussion
The latest freeze has renewed calls to limit how quickly a borrower can build a position, instead of relying solely on a total market ceiling and a response after an alert.
A total cap restricts the amount outstanding. A limit on the flow of new activity would restrict how much can enter during a specified period, even when there is room below that ceiling.
Governance participant gnarvaja had proposed an automated supply-cap updater in April. The proposal would keep permitted supply close to existing deposits and expand it in small steps within a maximum set by governance.
Its suggested parameters included roughly 3% headroom and updates no more frequently than once an hour. The author said the system could be implemented through an external contract with risk-administration permissions, without changing Aave V3’s core contracts.
The proposal would leave governance in control of the maximum and the updater’s settings. A security council would retain the ability to pause the system.
A response from Robby Greenfield supported tighter limits but raised questions about asset-specific settings and feedback loops. He said deposit activity could repeatedly fill a cap that automatically expands, so adversarial behaviour needed simulation before an implementation proposal.
The October incident thread brought the April proposal back into the discussion. Its author said the earlier suggestion had been ignored and asked why markets had to be frozen again.
Tyler Loewen’s Rivus Guard proposal, published in 2024, describes another automated controller for supply and borrow caps. It separates governance-set maximums from softer limits on how much totals can change over time.
The design leaves risk managers responsible for hard caps. Controllers would apply bounds to changes calculated by separate components, allowing daily adjustments without resetting the overarching maximum each time.
Loewen presented the softer limits as a way to restrict sudden changes in flows. The proposal is a design for additional controls, not evidence that those controls were operating during the Binance incident.
At 01:01 UTC on October 31, Loewen returned to the current incident discussion. He said talks about his controller had stalled over compensation and offered to drop his payment request and revisit the system without charge.
What to watch
Collateral permissions: Aave’s incident thread will carry the conditions for restoring borrowing power to the 9 restricted assets.
Borrowing capacity: Reserve caps and outstanding debt will show how much room remains for new loans after reopening.
Interest-rate settings: LlamaRisk said temporary changes would be reviewed as market conditions normalised.
Attacker position: Debt, collateral and any liquidation or repayment must be assessed at their own timestamps.
Automated controls: Governance proposals, votes and deployed contracts will show whether the renewed discussion produces an implemented limit.
FAQ
Was Aave hacked in the Binance incident?
Aave said its contracts were unaffected. A Binance-linked wallet used the lending system to borrow WETH against USDT before precautionary restrictions took effect.
How much did the attacker borrow from Aave?
LlamaRisk reported borrowing of 19,762 WETH, worth approximately $44.6 million at the time, against a deposit of 60 million USDT. Its initial account said the position remained overcollateralised.
Could users withdraw during the Aave freeze?
Withdrawals remained enabled, subject to available liquidity and collateral requirements. New deposits and borrowing were stopped, while repayments and liquidations continued.
Why did borrowing rates rise during the freeze?
Supplier withdrawals left more of the remaining pool borrowed, pushing utilisation higher. LlamaRisk said the elevated rates could not attract new deposits while supply was frozen, so it executed temporary rate changes.
Have Aave’s markets reopened?
Aave announced on October 30 that V3 and V4 markets were unfrozen. Nine assets retained restrictions on the borrowing power they provide as collateral.
Does zero LTV mean my tokens are worthless?
No. LTV describes borrowing power, not market price. V3’s retained restrictions do not change existing liquidation thresholds, while V4 has separate rules for when a changed collateral configuration applies.
Do the lower borrowing caps mean Aave lost $3.83 billion?
No. LlamaRisk’s figures describe unused borrowing capacity removed from V3 and V4. They are not an estimate of stolen funds or a write-off of existing deposits.
Would automated caps have prevented the attacker’s loan?
The proposals describe ways to limit rapid activity, but the reviewed evidence does not establish the result they would have produced for this particular transaction. This is educational analysis, not investment advice.
Disclaimer: This article is for information and educational purposes only and does not constitute financial or investment advice. Figures reflect regulatory filings and reporting available at the time of writing and change with each disclosure. Nothing here is a recommendation to buy, sell, or hold any security or asset. Always do your own research. Information is accurate as of October 31, 2026.
