LlamaRisk, the risk manager overseeing lending protocol Aave, plans to introduce time-based borrowing limits for the platform. Speaking on the Uneasy Money podcast, CEO Adam Sadowski explained that the new mechanism will cap how much users can borrow from Aave within specified time frames. The announcement came fewer than seven days after a wallet involved in the Binance hot-wallet exploit used USDT as collateral on Aave and borrowed ether against it.
Response to Exploit and Ongoing Precautions
The wallet, which Etherscan identified as part of the Binance hot-wallet exploit, posted more than $35 million worth of USDT as collateral on Aave V3’s Ethereum Core market. As of November 5, onchain data indicated that the wallet still owed approximately 9,885 WETH, backed by about 35.57 million USDT, resulting in a health factor near 1.24.
This debt represents about half the 19,762 WETH—valued at $44.6 million at the time—originally borrowed against 60 million USDT across three minutes on October 29, according to LlamaRisk. Eventually, a liquidator repaid 9,884 WETH and claimed roughly 24.38 million USDT in collateral when the position started liquidating. Sadowski noted that the position had “started to enter liquidation.” Calculations based on the health factor suggest that ether would need to rise another 24% for the position to face liquidation again.
Guardians of Aave, using a multi-signature system, temporarily froze the markets about 40 minutes after the exploit took place. Despite the freeze, liquidators continued operating, borrowers repaid loans, and users could withdraw collateral. However, new deposits and borrowings were temporarily disabled. The Ethereum Core market has since reopened, although USDT and USDC collateral does not provide new borrowing power since the loan-to-value ratio for those assets stands at zero.
LlamaRisk’s Solution: Automated Risk Management
Sadowski outlined ongoing efforts to automate risk measures on Aave. He introduced risk oracles as the central idea—a system that interprets real-world data and triggers automated actions on the protocol. For the upcoming feature, these oracles will enforce borrowing caps within certain periods, although specifics like limit sizes and intervals were left unspecified.
LlamaRisk is developing time-based restrictions on borrowing capacity by leveraging risk oracles, aiming to reduce the risk of rapid, large-scale exploits.
Aave’s protocol governance involves guardians, who operate through a four-of-seven multi-signature arrangement for major interventions, while risk stewards are managed collaboratively by LlamaRisk and Aave Labs. Sadowski emphasized the need to transition from manual oversight to automation due to the increasing presence of AI agents identifying and exploiting vulnerabilities.
He referenced Chainlink’s Runtime Environment (CRE), a framework enabling risk managers to execute their own logic and custom automation on decentralized systems. Currently, LlamaRisk’s risk oracle for Pendle principal tokens (PT) on Aave already operates on CRE, and the same approach is being planned for other risk measures such as supply and borrow caps.
Mini dictionary: Chainlink Runtime Environment (CRE): A secure execution environment that allows decentralized applications like oracles and risk management tools to run custom code on blockchain networks, enhancing automation and real-time responsiveness in DeFi.
Community Debate Over Implementation
While forum posts from LlamaRisk had not detailed time-based borrowing restrictions, community members discussed flow limits and automated supply-cap updates. User gnarvaja referenced his earlier proposal for an automated supply-cap mechanism.
Kain Warwick, founder of synthetic asset protocol Synthetix and co-host of the podcast, argued that per-account limits can be bypassed by splitting borrowing among multiple wallets. Sadowski responded by acknowledging inherent trade-offs in risk management, stating that “the Aave DAO sets the frequency and the parameters delegated operators can adjust.” Warwick also warned of significant unintended consequences when making fundamental changes to DeFi systems.
Among the ongoing concerns discussed, Sadowski highlighted the challenge of staying ahead of increasingly sophisticated actors as DeFi automation and AI technologies evolve.
