Aave Moves to Deprecate More Assets as Protocol Cleanup Continues

This article originally appeared in Galaxy Research's weekly newsletter. Subscribe to get timely insights delivered to your inbox every Friday morning.

WHAT HAPPENED

This week LlamaRisk, the primary risk service provider to the Aave DAO, set forth an Aave Request for Comment (ARFC) to begin freezing, migrating, and offboarding a collection of assets from Aave V3 Core, the top DeFi protocol’s most widely used version.

The move would be a continuation of a larger campaign by the protocol to tighten its operational risk in the wake of the KelpDAO exploit from April [1][2]. The cleanup effort has also included introducing a framework for assessing protocol risk, deprecating assets with little adoption and implementing updated oracle designs for tail assets on the protocol.

Now, LlamaRisk is calling to purge V3 of assets with the following characteristics:

• Restaking exposure: assets whose backing is delegated to secure other services or networks. rsETH is the main example on the list.

• Leveraged wrapping strategies: assets where most of the backing is itself a leveraged position in another asset, so the collateral and its underlying move together under stress.

• Derivatives strategies in the backing: staked or wrapped assets where part of the backing goes into derivatives beyond simple hedging, such as covered-call writing.

• Custom non-custodial bridge designs: wrapped assets whose backing is held by a purpose-built signer or enclave network rather than an established custodian.

• Reserves with limited demand: positions too small to justify ongoing maintenance, regardless of asset quality.

• Issuer risk: issuers with insufficient operational and security resourcing relative to the size of their exposure on Core, as judged by the Aave Risk Framework.

The significance of this move rests on the sheer size of the assets recommended to be offboarded. At the time of LlamaRisk’s governance post, it covered $1.33 billion of collateral assets and $1.03 billion of borrowed assets (roughly 8% of outstanding borrows on V3 Core). This potentially poses monumental operational efforts to get the assets migrated and offboarded and could crimp Aave revenue if not migrated effectively. Most of the size comes from assets used in looping strategies (i.e. liquid staking and liquid restaking tokens used to gain leverage to Ethereum’s staking rate). The proposal primarily targets ETH liquid staking/restaking tokens on a USD basis for removal. As demonstrated by Galaxy Research, the strategies in which these asset types are present pose acute risk to Aave in that they carry high amounts of leverage by users borrowing close to their liquidation thresholds. As a result, even small deviations in price as a result of issuer shortcomings can have large implications for Aave.

It’s worth noting that Aave is not rejecting these assets outright. The overarching plan is to make V3 Core a place for pooled blue-chip asset lending and borrowing, while V4, which has greater flexibility around risk siloing, can house a fuller spectrum of assets.

OUR TAKE

Net, this would be a positive step for Aave, and DeFi at large, to take. Aave is now building a more formal view on how assets are constructed and setting risk parameters that align with it, using past failures as a guide. The risk doesn’t fully disappear, however. The attack surface for the assets that remain active will still exist, and attackers will keep finding new angles of exploitation, possibly using AI assistance to break through.

This is why collateral listing is a security decision. Lending markets have been a popular route for attackers to turn forged and stolen assets into something they can keep at market value, by converting collateral into liquid assets instantly, at oracle prices, with no real slippage outside of loan-to-value parameters (and the assets they pilfer come at the expense of protocol liquidity providers). In the KelpDAO exploit, attackers minted approximately 116,500 unbacked rsETH tokens and used them as collateral to drain real wrapped ether from Aave’s pools, leaving roughly $200 million in bad debt. In the October Binance exploit, the attackers held USDT that Tether could freeze, so they deposited $60 million of USDT into Aave V3 and borrowed $44.6 million of wrapped ether around 40 minutes before the first freeze. That position stayed overcollateralized, but suppliers’ WETH ended up with the attacker, and recovery now depends on collateral that blacklisting the attacker’s wallet could not automatically retrieve from the pool.

From this perspective, the updates being proposed to Aave’s collateral set are about protecting liquidity providers (LPs). For example, on Aave V3 Core every WETH and stablecoin supplier underwrites every listed collateral asset at the same pooled rate, while the extra yield from riskier collateral mostly goes to the borrowers looping it rather than to the suppliers carrying the risk. Narrowing V3 Core to a tighter set of blue-chip collateral would rebalance this trade. LPs would take less tail risk for the same return and attackers would have fewer routes to liquidatingtheir positions into desirable, censorship-resistant assets.

What makes this repositioning possible is V4. Before Aave’s latest version launched, its choice was binary: list an asset and have every supplier underwrite it or turn it away and lose the business. V4 removes that tradeoff. Each collateral asset sits in a “spoke” with its own caps, and borrowers pay a premium in proportion to their collateral risk, so suppliers are paid for what they underwrite and exposure to any one asset is bounded and becomes a more explicit choice. That frees V3 Core to concentrate on blue chips while collateral assets the protocol views as riskier can be siloed off and be more compensatory to LPs.

The open question around the proposal is migration, and the risk stays in the meantime. A freeze would stop these markets from growing and cut off new collateral as a route for attackers, but it wouldn’t remove the exposure for LPs already on V3 Core. That will only disappear when the borrows backed by these assets are repaid or moved. rsETH has been frozen since April and still backs over $600M of WETH debt, so this process is likely to be slow, and suppliers will carry the legacy risk until it’s done. Notably, LlamaRisk proposed reflecting the Emode configurations of collateral assets being deprecated across those that will remain on the platform (e.g. rsETH’s emode configs will apply to weETH, wstETH, and cbBTC), so borrowers and loopers can swap out their collateral assets to maintain a loan with the same configuration.

The plan would work optimally only if V4 absorbs the demand V3 Core is pushing out, or if borrowers swap out deprecated collateral assets for retained ones. If borrowers move to V4 or swap into retained Core collateral, Aave will keep the business and shed the risk. If they leave, Aave will give up revenue to competitors willing to take the risk (a scenario which could also be read as the premium of taking risk being removed from protocol economics).

LlamaRisk is betting on the first outcome and moving slowly to get there: positions are frozen rather than forced out, and harsher levers, like higher borrow rates on the leveraged loops that get netted out through incentive programs for users not using collateral on the chopping block, are held back as a last resort.

Nonetheless, slowness is a reasonable price given the magnitude of the plan and the possible long-term benefits it should bring to V3 Core and the Aave product consortium generally.