This article was compiled and organized by BlockWeeks
Aave DAO's primary risk service provider LlamaRisk this week submitted an Aave Request for Comment (ARFC) proposal, planning to initiate a freeze, migration, and phase-out process for a batch of assets on Aave V3 Core. V3 Core is the most widely used version of this leading DeFi protocol.
This move is part of Aave's broader ongoing effort to tighten operational risk following the April KelpDAO attack. Previous cleanups have already included introducing a protocol risk assessment framework, delisting low-adoption assets, and designing an updated oracle solution for long-tail assets.
Six Phase-Out Criteria
LlamaRisk listed a total of six categories of phase-out conditions this time:
- Restaking exposure: The asset's reserves are re-delegated to secure other services or networks; rsETH is the primary example on this list.
- Leveraged wrapper strategies: The majority of the asset's reserves are themselves leveraged positions on another asset, so under stress scenarios the collateral and the underlying price decline in the same direction.
- Reserves containing derivative strategies: Part of the staked or wrapped asset's reserves are deployed into derivatives beyond simple hedging, such as covered call selling.
- Custom non-custodial bridge designs: The wrapped asset's reserves are held by a purpose-built signer or enclave network rather than by an established custodian.
- Reserves with limited demand: The position size is too small to support ongoing maintenance costs regardless of the asset's underlying quality.
- Issuer risk: Assessed under Aave's risk framework, the issuer's operational and security resource commitment is disproportionate to the scale of its exposure on Core.
Scale: $1.33 Billion in Collateral and $1.03 Billion in Borrows
The weight of this proposal lies in the volume of the assets recommended for phase-out. At the time LlamaRisk published the governance post, it involved $1.33 billion in collateral assets and $1.03 billion in borrowed assets, accounting for approximately 8% of V3 Core's outstanding borrows. Completing the migration and phase-out of these assets entails an enormous operational workload; if the migration proves ineffective, it could also erode Aave's revenue.
The vast majority of the volume comes from assets used in looping leverage strategies, i.e., amplifying exposure to Ethereum staking yield through liquid staking tokens (LSTs) and liquid restaking tokens (LRTs). The proposal primarily targets USD-denominated ETH liquid staking and restaking tokens. Research from Galaxy Research shows that the strategies these assets participate in pose acute risks to Aave: users' leverage levels are high, and borrow sizes sit close to liquidation thresholds, so even a small price deviation triggered by issuer issues could have a major impact on Aave.
Notably, Aave is not seeking to reject such assets outright. The overall plan is to make V3 Core a pooled lending venue for blue-chip assets, while V4, which offers greater risk isolation flexibility, can accommodate a fuller spectrum of assets.
Collateral Listing Is Fundamentally a Security Decision
For Aave and DeFi as a whole, this is a positive adjustment. Aave is establishing a more formal set of criteria for judging asset structure and setting risk parameters to match accordingly, guided by past failure cases. But risk will not simply disappear: assets still on the platform retain their attack surface, and attackers will keep looking for new angles of exploitation, potentially even using AI to breach defenses.
This is precisely why collateral listing must be treated as a security decision. Lending markets have long been a popular channel for attackers to monetize forged and stolen assets: they can instantly swap collateral into liquid assets, transact at oracle prices with virtually no real slippage beyond the loan-to-value parameter, and the plundered assets ultimately end up being paid for by the protocol's liquidity providers.
In the KelpDAO attack, the attacker minted approximately 116,500 unbacked rsETH and used it as collateral to drain real wrapped Ethereum from Aave pools, leaving roughly $200 million in bad debt. And in the October Binance incident, the attacker's USDT holdings fell within Tether's freezable scope, so about 40 minutes before the first freeze, they deposited $60 million USDT into Aave V3 and borrowed $44.6 million in wrapped Ethereum. That position remained overcollateralized throughout, but the suppliers' WETH ultimately ended up in the attacker's hands, and recovery now depends on that portion of collateral in the pool—simply blacklisting the attacker's wallet does not automatically retrieve those assets from the pool.
Whose Risk Are LPs Actually Bearing
From this perspective, the core of this adjustment to Aave's collateral set is protecting liquidity providers (LPs). For example, on Aave V3 Core, every WETH and stablecoin supplier backstops all listed collateral assets at the same pooled interest rate; yet the extra yield generated by high-risk collateral mostly flows to looping borrowers rather than to the suppliers bearing the risk. Narrowing V3 Core to a stricter blue-chip collateral set can rebalance this trade structure: LPs bear less tail risk for the same return, and attackers have fewer paths to liquidate their positions into desirable, censorship-resistant assets.
V4 Provides a Solution That Was Not Previously Available
What makes this repositioning possible is V4. Before the latest version launched, Aave faced a binary choice: either list an asset and have all suppliers backstop it, or refuse and lose the business. V4 eliminates this trade-off. Each collateral asset is placed in its own independent "spoke" with its own cap, borrowers pay a premium proportional to collateral risk, and suppliers are accordingly compensated for the risk they underwrite; collateral assets the protocol deems riskier can be isolated and LPs compensated more.
Migration Is the Biggest Uncertainty, and Risk Remains on Site
The real suspense of the proposal lies in migration. Freezing will prevent these markets from expanding further and cut off attackers' path of using new collateral as a channel, but it will not eliminate the exposure of LPs already in V3 Core. Exposure will only truly disappear once the borrows backed by these assets are repaid or transferred. rsETH has been frozen since April but still backs over $600 million in WETH debt, so this process is likely to be quite slow, and suppliers will continue to bear legacy risk during the transition period.
Notably, LlamaRisk proposes mapping the Emode configurations of the phased-out collateral assets onto assets that will remain on the platform. For example, rsETH's emode configuration would apply to weETH, wstETH, and cbBTC, so that borrowers and looping users can switch collateral assets while maintaining the same loan configuration.
For this plan to achieve optimal results, the premise is that V4 can absorb the demand squeezed out of V3 Core, or that borrowers swap the abandoned collateral assets for retained ones. If borrowers move to V4 or rotate into retained Core collateral, Aave both retains the business and sheds the risk; if they choose to leave, Aave is ceding revenue to competitors willing to take on the risk—a scenario that can also be interpreted as the "premium for bearing risk" being removed from the protocol's economics.
LlamaRisk is betting on the first outcome and proceeding at a slow pace: positions are frozen rather than forcibly liquidated; harsher measures (such as raising borrow rates on looping positions and then using incentive programs to net-hedge users outside the abandoned collateral) are kept as last resorts.
Conclusion
Nevertheless, given the scale of the plan and the long-term benefits it may bring to V3 Core and the entire Aave product suite, moving slowly is a reasonable price to pay.
