Curve LlamaLend

Protocol identification (as of 10 October 2026)

• Name/category: Curve LlamaLend (Curve Lending), an isolated, permissionless lending market system using ERC-4626 vaults, borrower controllers and LLAMMA soft-liquidation AMMs. v1 is deprecated; v2 is the current design.
• Website/docs: Curve’s frontend (Curve Finance) and Curve Knowledge Hub documentation; no separate, independently verified LlamaLend domain was identified. Launch: v1 launched 12 March 2024; v2 launched on Ethereum 13 July 2026 (after Optimism).
• Chains: Ethereum is confirmed for this review. The v2 registry also lists Optimism; historical v1 expansion included Arbitrum, Optimism and Fraxtal.
• Native token: No LlamaLend-specific token. CRV is Curve DAO’s governance/incentive token; crvUSD is an asset used by v1 markets, not LlamaLend’s native token. Ethereum contract identification / cross-check
• v2 LendFactory: 0x8f6B56EC5ddF1F2691a1059f1D3cd97Ac9EaB0bd
• Example v2 market 0: Vault 0x2b5a321c3cb1f33e1abecd047c2649d0b4c47eba; Controller 0xc77d97cf01737eb7ace46cab7cd9f60ec51a40c0; LLAMMA 0xbf6f64b741164c26023f97faaea8e02453c27442.
• Legacy v1 OneWayLendingFactory: 0xeA6876DDE9e3467564acBeE1Ed5bac88783205E0; lending AMM implementation 0x0ec8e0c868541df59ceD49B39CC930C3a8DbD93a. Addresses cross-check between Curve documentation and linked Etherscan records.
• Dune cross-check: Not verifiable as of 10 October 2026 (Dune MCP unavailable). Etherscan source-verification badges were also not independently confirmed because address-page retrieval returned 403; Not verifiable as of 10 October 2026. Fork lineage and security LlamaLend is not a conventional third-party fork; it is Curve’s lending implementation derived from Curve’s curve-stablecoin/crvUSD codebase. v1 reused LLAMMA unchanged and modified Controller logic for lending; v2 replaces duplicated monolithic controllers with Vyper modules, adds non-crvUSD pairs, borrow/supply caps, DAO-gated oracle configuration and improved vault accounting. v1 received StateMind/MixBytes reviews. For v2, Curve governance says ChainSecurity completed an audit, while current documentation still says audits are “in progress”—a material documentation contradiction. Fork-risk history: the Resupply hack exploited an inherited v1 vault-accounting/inflation weakness; this is a known analogous fork/integration failure, not evidence of malicious modification. No documented malicious code modification of LlamaLend itself was found.

Curve LlamaLend appears to be a mature, live product rather than a mere landing page: Curve’s docs describe LlamaLend v2 as permissionless isolated lending markets with ERC-4626 vaults, a LendFactory, a Configurator, and LLAMMA soft liquidations, and the docs include concrete deposit/withdraw workflows plus market-discovery and integration guidance. The documentation also states that LlamaLend v1 is deprecated and that v2 is the reference for new markets and integrations, which is consistent with an active production app rather than a template site. The user-facing functionality looks real: Curve documents supplying assets on Llamalend, and the docs explicitly say users can withdraw anytime as long as there is enough unborrowed liquidity in the pool.

The protocol docs also provide programmatic market discovery and contract interfaces, including the public Curve API for pool data, prices, and analytics, which indicates an open API exists for integrators. What is not verifiable from the available web evidence is live UI quality details such as broken links, fake metrics, or template-sign detection; those require hands-on site inspection and were not confirmed here. Also not verifiable as of 2026-10-10: whether deposits/withdrawals are currently functioning on the specific front-end for every market, because the gathered sources are documentation and ecosystem references rather than a live UI session.

Overall assessment: high product maturity, active documentation, clear integration surface, and an open API for data/integration use.

Curve maintains an active bug bounty program for security disclosures. The documented scope covers issues that can cause substantial loss of funds, broken liveness conditions, or irreversible loss of funds. Eligibility requires being the first reporter, being able to verify a signature from the same address, and providing enough information about the vulnerability.

Exclusions include already known vulnerabilities and front-end issues that do not lead to smart contract vulnerabilities. The published payout matrix tops out at $250,000 for high-severity, almost-certain issues, with lower payouts of $50,000 / $10,000 / $1,000 / $250 depending on severity and likelihood. Results visible from the gathered sources include prior bounty-related payouts referenced by third parties, but no authoritative bug-bounty disclosure list for LlamaLend specifically was found in the collected sources.

Not verifiable as of 2026-10-10.

Scope: Curve LlamaLend on Ethereum includes legacy v1 markets and newer v2 isolated markets. Dune/on-chain verification was unavailable; therefore current market inventory, balances, collateral composition, utilization, and exposure percentages are Not verifiable as of October 10, 2026. External dependencies / counterparties

• Curve AMM/LLAMMA: Each market depends on its own LLAMMA for soft liquidation and conversion between collateral and borrowed assets. Liquidation execution depends on arbitrageurs/liquidators and sufficient market liquidity; these are permissionless, not guaranteed counterparties.
• Oracles: Material oracle risk. v2 accepts market-specific custom oracles; pricing must resist manipulation, stale data, discontinuities, and liquidity-source failure. Curve-pool oracles may rely on shallow pools. Governance previously identified Ethereum LlamaLend markets with unsafe oracle pools below $100,000 TVL, including ezETH and swBTC markets.
• Governance/admin: Parameter, oracle, borrow-cap, and fee changes are controlled through the permissioned Configurator; default administration is assigned to Curve DAO ownership agents, creating governance/key-management dependence.
• Stablecoins, LSTs/restaking, and issuer risk: v2 permits arbitrary borrowed/collateral token pairs and supported Curve LP tokens. Consequently, exposure may include crvUSD, third-party stablecoins, LSTs/restaking tokens, LP tokens, or other issuer-dependent assets. The actual Ethereum allocation is Not verifiable as of October 10, 2026.
• Bridges/custodians/CEX/MMs/RWA SPVs: Ethereum-only deployment has no necessary bridge dependency for core settlement. No custody or RWA-SPV dependency is established in the reviewed sources. CEX/MM exposure is indirect and market-specific; Not verifiable as of October 10, 2026. Failure scenarios: oracle manipulation or staleness can misprice collateral; collateral/stablecoin/LST depeg can trigger LLAMMA conversion and bad debt; thin liquidity, gas congestion, or absent liquidators can impair liquidation; governance compromise or faulty parameter changes can expand caps or replace oracles. Structured fields: dependency_failure_active: null max_exposure_pct: null

Curve LlamaLend is non-custodial: users keep ownership of their assets, and the protocol uses isolated markets with separate risk parameters for each collateral/borrow pair. On the architecture described in Curve’s docs, lenders deposit into an ERC-4626 Vault, borrowers interact through a Controller, and collateral is held by the LLAMMA/AMM for soft-liquidation mechanics; Curve also states collateral is not re-hypothecated. This means custody is organized at the smart-contract level rather than by a centralized custodian, with assets segregated by market so exposure in one market does not contaminate others.

Key-management organization — Curve LlamaLend (Ethereum)

• Primary control layer: LlamaLend v2 markets are administered through a permissioned Configurator, which can change controller, AMM, oracle, monetary-policy, lending, and fee parameters. Users and market deployers do not hold these privileges by default.
• Default administrator: Curve’s documentation states that the Ethereum Configurator’s default_admin is the Curve DAO Ownership Agent (0x4090…9968, abbreviated in the documentation). Activating a newly deployed market—whose borrow cap initially equals zero—therefore requires a Curve DAO ownership vote unless governance delegates administration to a market-specific address.
• Delegated administration: The default administrator may assign a custom administrator per LlamaLend controller. That address can manage the assigned market’s permitted settings, while changing or removing the delegation remains restricted to the default administrator. Setting the custom admin to the zero address returns control to the default admin.
• Governance execution: Curve governance uses an Aragon-based voting system. Ethereum votes are executed through separate DAO agents, including Ownership, Parameter, and Emergency Agents; the Ownership Agent is used for critical protocol-control actions. LlamaLend’s documented default-admin path specifically points to the Ownership Agent.
• Key custody / signer policy: The reviewed sources identify the controlling smart-contract addresses and governance path, but do not establish the current private-key custody model, signer identities, multisig threshold, hardware-wallet policy, or operational key-rotation procedures behind the DAO agents. Not verifiable as of October 10, 2026. Risk implication: LlamaLend’s market administration is not controlled by the market deployer; it is DAO-gated by default, with possible per-market delegation. The residual governance risk depends on the security and decentralization of the Curve DAO agent execution path and any delegated custom-admin addresses.

Scope: Ethereum LlamaLend v2; legacy v1 deployments are separate. As of October 10, 2026. Addresses / architecture

• LendFactory: 0x8f6B56EC5ddF1F2691a1059f1D3cd97Ac9EaB0bd
• Configurator: 0x6065858d0eF0AA240DFdf6f1A0B2ae34B41f49bC
• Blueprints: AMM 0xc8AC252738E1Ece3f69CF77649C266c4E893cf41; LendController 0x47b6dF6494aD62474cDF365B90a56C648778A75d; Vault 0x2c3822264dcbd18d910C7834b1De8A70f368375b; ControllerView 0x7259…B2C (full address not independently recovered).
• Each market is deployed as Vault + LendController + LLAMMA AMM, created from blueprints; Configurator governs market parameters. This is not an EIP-1967 proxy architecture. The contracts are therefore not generally implementation-upgradeable, but market oracle/configuration is changeable through the Configurator. Admin powers / risk
• Factory owner: can pause new market creation, change supply caps and fee receivers, and transfer ownership. Existing markets continue operating when the factory is paused.
• Configurator default admin: can assign per-market admins and change oracle, monetary policy, AMM fee, discounts, borrow cap and admin percentage. Documentation reports Ethereum default admin as Curve’s Ownership Agent 0x4090…9968; exact signer threshold/timelock is Not verifiable as of October 10, 2026 because Dune was unavailable.
• Vault withdrawals are user/ERC4626-driven and limited by available liquidity; no documented admin sweep function was found. Users can generally exit while liquidity remains available, but an adverse oracle/parameter change can create liquidation or insolvency losses.
• Compromised governance/admin keys: market-specific oracle manipulation, extreme fees/discounts/rates, borrow-cap changes, fee redirection, and market freeze via caps are credible worst cases; arbitrary instantaneous withdrawal of all lender funds is not evidenced.
• Timelock delay, role renunciation, proxy-admin events, and current admin event history: Not verifiable as of October 10, 2026. Public v2 audits were stated to be in progress/not released. Diagram DAO/Ownership Agent → Configurator → {Controller, AMM, Oracle/rate parameters} Factory owner → LendFactory → {market creation, supply caps, fee receivers} Lenders → Vault ↔ Controller ↔ LLAMMA AMM ↔ Collateral/borrowed tokens Fields: admin_can_drain=false; audited_deployment=false; upgradeable=false; unresolved_critical=null; unresolved_high=null

Published Curve Lending/LlamaLend audit report.

Published Curve LlamaLend/Lending audit report.

Founder and team. Curve LlamaLend is a Curve Finance product, not an independently incorporated or separately staffed protocol. The identifiable founder is Michael (Mikhail) Egorov, who founded Curve in 2020 and previously founded/co-founded NuCypher, where he served as CTO from 2015–2020. Public reporting describes Curve’s early team as very small—approximately five people, including two developers—and based in Switzerland.

A complete, independently verified LlamaLend team roster is Not verifiable as of October 10, 2026. Public versus anonymous. Egorov is publicly identifiable, but most protocol contributors are not presented as a conventional, fully disclosed corporate team. LlamaLend’s operational identity is therefore best characterized as founder-led and partially pseudonymous, with governance and open-source development substituting for normal corporate transparency. A named legal entity, beneficial-ownership record, and independently verified operating-office address are Not verifiable as of October 10, 2026. Prior projects and outcomes. NuCypher gives Egorov relevant cryptography and protocol-engineering experience.

Curve itself is a genuine, widely used on-chain DeFi protocol rather than a merely informational website; however, its history includes a major July 2023 exploit linked to vulnerable Vyper compiler versions, with approximately $62 million initially reported stolen and partial recovery afterward. LlamaLend launched after that incident, and a pre-launch Statemind review reportedly identified one critical, one high, and four medium issues that were reportedly remediated before deployment. LlamaLend risk track record. Reported incidents include approximately $10 million of LlamaLend bad debt associated with Egorov’s highly leveraged CRV positions, plus later market-specific oracle and bad-debt events. These are material governance, risk-management, and market-design concerns even where they do not demonstrate a direct compromise of LlamaLend’s core contracts. > Reality check: Switzerland/Zug or Zürich is repeatedly reported as Curve’s base, but the sources do not establish a verified staffed office, legal domicile, or onshore/offshore structure.

Accordingly, “real office,” jurisdiction, and business substance are Not verifiable as of October 10, 2026. The evidence supports a real operating DeFi project with public founder history, not proof of a conventional, transparently disclosed company. Sources used: RootData, ForkLog, IQ.wiki, Avoid.net.

Reputation assessment — Curve LlamaLend (Ethereum): mixed, with material unresolved lending-market risk. The protocol benefits from Curve’s established DeFi profile, public governance, an active bug-bounty program, and audits involving firms including StateMind, ChainSecurity, Trail of Bits, MixBytes, and Quantstamp. However, the evidence also shows repeated security and solvency concerns rather than a clean operating record.

• Founders/investors: Michael Egorov is Curve’s founder and was publicly associated with a major 2024 deleveraging episode. No independently verified institutional-investor roster for LlamaLend was identified. Not verifiable as of October 10, 2026.
• Audits: StateMind reportedly identified one critical, one high, and four medium findings in a pre-launch LlamaLend audit, described as remediated before deployment. ChainSecurity also identified a critical crvUSD FlashLender drain scenario in a broader Curve review; governance subsequently approved corrective changes.
• Incidents and criticisms: The Ethereum sDOLA LlamaLend market suffered a March 2026 price-manipulation/oracle exploit with approximately $239,000–$240,000 in losses. Reports attribute the weakness to market configuration and price manipulation, including concentrated sDOLA liquidity. In 2025, the CRV-long LlamaLend market reportedly accumulated about $700,000 of bad debt during a sharp market liquidation, leaving the lending vault materially undercollateralized.
• Bad debt and founder exposure: Following the 2024 UwU Lend exploit, CRV collateral liquidations reportedly created roughly $10 million of LlamaLend bad debt; Egorov stated that he repaid it. This episode raised concerns about founder leverage, collateral concentration, and contagion between external protocols and LlamaLend.
• Fraud, rug, insolvency: No reliable evidence in the retrieved sources establishes fraud, an intentional rug pull, or protocol-wide insolvency. Market-level bad debt and lender losses are documented allegations/events, not proof of fraudulent intent.
• Legal, regulatory, sanctions: Not verifiable as of October 10, 2026. Important limitation: Ethereum on-chain verification was unavailable in this run; therefore TVL, exposure, repayment status, loss figures, and current solvency cannot be independently confirmed. Some retrieved sources are secondary or risk-analysis sites, so their incident figures should be treated as reported—not on-chain verified.

Economic model — Curve LlamaLend, Ethereum Strategy / asset flows. Isolated, one-way lending markets: lenders deposit the *borrowed asset* into an ERC-4626 vault; borrowers deposit one collateral asset and borrow against it. Lender yield is primarily borrower interest, accrued through rising vault share price. V2 permits flexible pairs beyond crvUSD, including supported Curve LP, yield-bearing and principal-token collateral. Yield quality. Base yield is organic only when generated by borrower interest/utilization.

CRV liquidity-mining callbacks can add subsidized yield to borrowers; any displayed incentives should therefore be separated from interest income. Organic-yield percentage: Not verifiable as of October 10, 2026. APY history, volatility and sustainability: Not verifiable as of October 10, 2026; no reliable independent time series was found. Risk posture. This is not inherently market-neutral.

Lenders have isolated credit, liquidity, oracle and collateral-liquidation exposure; borrowers are directional to collateral versus borrowed asset. LLAMMA distributes collateral across price bands and gradually converts it into the borrowed asset as price falls. Recovery can reverse some conversion direction, but not prior conversion losses; hard liquidation remains possible at zero health. Leverage / external exposure. Native markets are isolated and do not rehypothecate collateral in v1.

V2 includes a LeverageZap and supports productive collateral, so looping/amplification is possible, but the live maximum leverage by market is Not verifiable as of October 10, 2026. No restaking or external lending dependency is intrinsic to the core market; LP collateral may retain Curve trading-fee exposure. Lock-ups / withdrawals / limits. No fixed maturity is documented. Lenders use ERC-4626 deposit/withdraw/redeem mechanics, subject to available liquidity and supply limits.

Borrowing is controlled by per-market borrow caps; Ethereum V2 markets initially launched with zero caps, requiring governance activation. Oracle, monetary policy, AMM fee, discounts, admin percentage and fee receiver are market-specific. Revenue. Potential protocol revenue is configurable admin fees; actual realized revenue and distribution are Not verifiable as of October 10, 2026. TVL / trend. Dune verification is unavailable in this run. Curve LlamaLend TVL total, by product, Ethereum share and trend: Not verifiable as of October 10, 2026. Contradiction / name-collision alert. The surfaced DeFiLlama llamalend page describes an NFT-lending protocol, not Curve LlamaLend; it is excluded from Curve TVL analysis.

Scope: Curve LlamaLend on Ethereum, primarily the current v2 deployment; v1 is deprecated. LlamaLend is an isolated lending-market system, not a reserve-backed stablecoin, and its market assets remain in market-specific ERC-4626 vaults, controllers, and LLAMMA AMMs rather than a single protocol treasury. Known contract infrastructure (Ethereum): LendFactory 0x8f6B56EC5ddF1F2691a1059f1D3cd97Ac9EaB0bd; example market 0: Vault 0x2b5a321c3cb1f33e1abecd047c2649d0b4c47eba, Controller 0xc77d97cf01737eb7ace46cab7cd9f60ec51a40c0, AMM 0xbf6f64b741164c26023f97faaea8e02453c27442, and ControllerView 0xcda563f85388e621e7d810387e5afdac5d395e2b. These are market contracts, not confirmed treasury addresses. Composition/custody: Lenders deposit the borrowed asset into each market’s ERC-4626 vault; funds lent to borrowers are unavailable for immediate withdrawal until repayment.

Collateral and converted assets are held across controller/LLAMMA positions. This is segregated market liquidity, not a discretionary reserve. Control: The LendFactory owner controls the default and custom fee receivers. Market administration—including borrow caps, admin percentage, oracle and monetary-policy settings—is generally assigned to Curve DAO ownership agents unless delegated.

The factory can route fees to a DAO, issuer, or curator. Reserve policy / attestations: No protocol-level reserve policy, treasury composition report, proof-of-reserves, or independent attestation for LlamaLend was identified in the reviewed sources. Public documentation describes fee routing and market controls, not a separately managed reserve. Audits for v2 were stated to be in progress and not publicly released on the documentation page. On-chain balances via Dune: Not verifiable as of October 10, 2026.

Dune was unavailable for this run; therefore no reserve balance, treasury balance, market TVL, or debt figure is asserted. No liabilities figure is inferred from aggregator data.

Conclusion: Curve LlamaLend has no protocol-specific native token. Do not attribute CRV tokenomics directly to LlamaLend. CRV is the native token of the broader Curve ecosystem. Relevant ecosystem token — CRV (Ethereum): contract 0xD533a949740bb3306d119CC777fa900bA034cd52.

Current third-party figures are approximately 1.573B circulating, 2.425B total supply, 3.03B maximum supply, $568M market cap, and $876M FDV; these are analytics-platform figures, not on-chain-verified results. Utility/governance: CRV rewards liquidity providers, including lending-market participants where a gauge/LM callback is approved; locking CRV as veCRV enables Curve governance, gauge voting, fee sharing, and reward boosting. LlamaLend v2 markets can stream CRV only after Curve DAO approval and gauge weight. Revenue, buybacks, burns, staking: LlamaLend revenue is generated through borrower interest and configurable market/admin fees; documentation permits fee receivers such as a DAO, issuer, or curator. No LlamaLend-specific CRV buyback, burn, or staking mechanism is documented. veCRV is governance locking, not LlamaLend staking. Emissions/unlocks/allocations: CRV community emissions decline 16% each August; initial team, investor, employee, early-user, and reserve vesting reportedly completed by August 2024, with remaining emissions directed primarily to liquidity providers over roughly 200 years.

Whether each announced unlock occurred on-chain, and exact allocations, are Not verifiable as of October 10, 2026 without Dune/on-chain execution checks. Controls and concentration: LlamaLend v2 administration is permissioned through Configurator contracts; Curve DAO ownership agents are documented as default administrators. Exact current admin, fee-switch, mint/blacklist authority, CRV top-holder concentration, insider wallets, and whether any blacklist exists are Not verifiable as of October 10, 2026. Liquidity/listings: LlamaLend itself has no token, hence no native DEX listings or liquidity depth. CRV’s venue-by-venue DEX depth is Not verifiable as of October 10, 2026 in this run.

Under a Bitcoin drop below $10,000, Curve LlamaLend on Ethereum would face an *extreme stress case* in which BTC-backed positions are very likely to enter the liquidation range and continue accruing losses through LLAMMA’s gradual conversion mechanism. Curve’s own documentation says LLAMMA does not liquidate at a single price; instead, collateral is progressively converted inside the liquidation range, and if health reaches 0% the loan is fully liquidated. The main protocol-level risk is that a prolonged, deep crash can turn soft liquidation into full collateral conversion, bad debt, and impaired lender withdrawals.

Curve has explicitly acknowledged that some permissionless Llamalend markets using volatile collateral accrued bad debt after sharp price moves and rapidly changing liquidity conditions, and that some lenders faced impaired withdrawals and prolonged uncertainty. Independent analysis also flags sustained multi-week declines as the worst case for LLAMMA because repeated conversions can leave borrowers worse off than a single liquidation. For a BTC price below $10,000, the relevant conclusion is not a precise percentage loss for the protocol, because that depends on each market’s parameters, current collateral mix, leverage, and liquidity depth, which are not verifiable as of 2026-10-10 from the provided sources.

What *is* clear is that BTC markets would be under severe pressure: borrowers would likely have little room to manage positions, lenders could face delayed or reduced recoveries, and any thin on-chain liquidity would increase the chance that conversions cannot keep pace with market moves. A concise risk view for Ethereum LlamaLend is:

• Borrowers: very high likelihood of entering liquidation protection, then partial or full loss of collateral if the drawdown persists.
• Lenders: elevated risk of impaired withdrawals and losses if bad debt forms in BTC markets.
• Protocol: stress on LLAMMA’s ability to convert collateral smoothly; the deepest failure mode is cascading bad debt rather than a simple one-price liquidation. Not verifiable as of 2026-10-10: exact BTC market exposure, current health distribution, and post-shock loss estimates for Curve LlamaLend on Ethereum.

For Curve LlamaLend on Ethereum, a 20% depeg of the largest collateral means the position can move into or through the liquidation range, triggering soft liquidation: collateral is gradually converted into the borrowed asset rather than being closed all at once. The protocol’s own docs state that liquidation is range-based, health can deteriorate to 0%, and once health reaches 0% the loan is fully liquidated. What this implies in stress terms: a 20% collateral drop is material but not automatically fatal; the actual loss depends on the market’s configured liquidation band, the position’s initial health, and whether the borrower can add collateral or repay before health reaches 0%.

Because LlamaLend is designed to keep risk contained within each market and not contaminate other markets, the primary impact is expected to be local to the affected market rather than system-wide. I cannot quantify the exact loss, bad debt, or TVL at risk for the largest collateral on Ethereum from the provided sources alone. Not verifiable as of 2026-10-10.

Scope limitation: The identity and exposure of Ethereum’s largest LlamaLend borrower are Not verifiable as of October 10, 2026 because Dune/on-chain verification is unavailable. “Insolvent” is therefore modeled as a borrower whose collateral value falls below accrued debt and who cannot repay. Stress scenario — top borrower defaults 1. Orderly liquidation — expected loss: low/zero for lenders.

• The borrower’s debt continues accruing in the market’s LendController.
• As collateral enters its LLAMMA bands, the AMM gradually converts collateral into the borrowed asset. Conversion fees, arbitrage discounts and volatility create a loss to the borrower.
• Once health reaches the hard-liquidation threshold, a liquidator calls LendController.liquidate(), repays debt and receives collateral subject to the liquidation discount. The borrower receives only any residual collateral.
• Who absorbs it: borrower absorbs liquidation losses; liquidator is compensated by the discount; lenders are made whole if recovered collateral covers debt, fees and slippage. 2. Bad-debt liquidation — expected loss: lender principal impairment.
• A fast price gap, oracle lag, thin collateral liquidity or adverse LLAMMA conversions can leave collateral proceeds below outstanding debt.
• The Controller closes/seizes the position, but cannot create missing borrowed tokens. The deficit remains inside that isolated market; it is not automatically transferred to other Curve markets.
• Who absorbs it: lenders through the ERC-4626 Vault’s reduced asset backing/share value and potentially constrained withdrawals. Curve’s lending audit explicitly identifies lenders—not the protocol—as bearing lending-market insolvency risk.
• Compensation: lenders retain interest earned and any residual collateral value only; no protocol-wide insurance or guaranteed DAO recapitalization is established by the cited architecture. Contagion path: borrower default → Controller health/liquidation → LLAMMA collateral sales → Vault asset shortfall → lender NAV/withdrawal impairment. Cross-market contagion is structurally limited by market isolation, but collateral-token price impact can affect external markets.

For Curve LlamaLend (Ethereum), a DAO/owner fraud stress scenario is not supported by the available evidence. The documented incidents in the provided sources are described as oracle-design failures, liquidations, bad debt, and third-party exploits, not as the DAO or owners stealing user funds or executing fraudulent withdrawals. The strongest governance-level evidence instead shows the DAO discussing recompensation for affected borrowers after the March 2026 sDOLA-long2 liquidation event, and explicitly framing the cause as a protocol vulnerability in the market design rather than borrower misconduct or owner fraud.

A Curve post also says the loss could be compensated by the DAO *if it approves*, which is the opposite of an established fraud finding. There are, however, governance and concentration risks that could worsen losses in a severe scenario: Curve’s own materials say LlamaLend markets are risk managed market by market and require DAO approval before debt can be built, while market design limitations can still produce hard liquidations and bad debt if oracle assumptions fail. The available sources do not show a verified case of the DAO or protocol owners committing fraud against LlamaLend users. Assessment: committed fraud by the DAO or owners = Not verifiable as of 2026-10-10.

For Curve LlamaLend, the primary lender yield source is borrowing interest; additional sources are CRV rewards and other token incentives. In a stress scenario where the primary yield source is negative over 30 days, the practical interpretation is that organic lending income has not covered the lender’s economic cost base for that period, so realized returns would depend on whether incentive emissions offset the shortfall. Curve’s docs do not state that the underlying lender supply rate can become mathematically negative; they describe supply yield as dynamic and tied to borrower interest and utilization, plus separate staking incentives.

So a "negative 30d" outcome is best treated as net yield compression to below zero after fees/costs, not a verifiable on-chain negative supply rate. Under stress, the main downside is that subsidy-dependent yields can expire or compress sharply; Curve’s own recap notes that when incentives end, effective APY drops from subsidized levels to much lower organic yields, and earlier campaigns even drove borrow rates deeply negative only while subsidies were active. Independent analysis of LlamaLend also flags incentive expiry as a failure path where depositor economics deteriorate materially.

For this question, the key risk assessment is:

• Income risk: primary yield can fall to near-zero or below your all-in hurdle once incentives fade.
• Subsidy risk: if CRV/other rewards are the main support, a 30-day negative primary source can be masked temporarily, but not sustainably.
• Stress implication: if utilization drops or borrow demand weakens, lender yield falls because supply rate is directly tied to borrower interest paid into the market. Not verifiable as of 2026-10-10: the actual Ethereum LlamaLend 30-day net yield for the selected market, because on-chain verification is unavailable in this run.

Scope: Ethereum, LlamaLend v2; v1 is deprecated. Contracts are controlled primarily by Curve DAO governance: the Ethereum Configurator’s default admin is the Curve Ownership Agent (0x4090…9968), and market-specific admins may be delegated by DAO action. The Configurator can change borrow caps, oracle, monetary policy, AMM fees, discounts, admin-fee percentage, and custom administrators. Proposal process / DAO reality. This is substantive rather than symbolic for documented LlamaLend parameters: veCRV holders with at least 2,500 veCRV may create immutable on-chain proposals; ownership votes require 30% quorum and 51% support, parameter votes 15% quorum and 60% support, with a seven-day voting period. Curve states LlamaLend proposals take approximately seven days from creation to execution. Emergency and funds. Curve has a separate Emergency Agent/Safe (0x4679…B1E0) with emergency powers, but its current signer identities and threshold are Not verifiable as of October 10, 2026 from the available sources.

The documented LlamaLend admin paths change parameters and factory/supply controls; no documented admin function directly transfers lender principal from the ERC-4626 vault. Therefore admin_can_drain=false for the documented design, subject to contract-level verification. DAO treasury funds are governance-controlled; a separate emergency proposal has authorized limited treasury transfers. Concentration / frontend / legal entity. Top veCRV holders and voting concentration via Dune: Not verifiable as of October 10, 2026 (Dune unavailable).

Frontend hosting control, developer legal entity, jurisdiction, registration number, directors, and ToS counterparty: Not verifiable as of October 10, 2026. The frontend/code is publicly maintained in Curve repositories, but repository ownership does not establish a legal operator.

Scope: Curve LlamaLend is a permissionless, non-custodial Ethereum lending protocol implemented through smart contracts; Curve documentation describes isolated one-way lending markets and no account-opening workflow. Entity / jurisdiction. The protocol itself is not shown to be a registered legal person. Governance is exercised by the Curve DAO through on-chain agents. Swiss Stake AG, an active Swiss company in Zug, is the principal identified developer/service provider for Curve; governance materials describe a *planned* Swiss Association wrapper, not an established legal owner. Terms, restrictions, KYC/AML. I found no LlamaLend-specific Terms of Service, restricted-jurisdiction policy, KYC/AML program, or user-contracting entity. Not verifiable as of October 10, 2026. Curve Pay’s KYC/AML and sanctions terms concern a separate custodial/payment product and should not be attributed to LlamaLend.

The smart-contract layer is permissionless; frontends, RPCs, stablecoin issuers, or integrators may impose separate screening or geographic restrictions. Classification. No regulator-specific classification of LlamaLend was identified. Functionally it is decentralized lending/credit infrastructure, but lending, securities, commodity, derivatives, money-transmission, or VASP treatment may vary by asset, activity, user location, and degree of control. This is a material legal-structure-versus-actual-risk gap: Swiss incorporation of a developer does not clearly allocate liability for DAO governance, deployed contracts, or independent users. Warnings, enforcement, courts, sanctions. U.S.

Treasury has cited Curve as a DeFi protocol in systemic/operational-risk discussion following the 2023 exploit, but this was not an enforcement action. Protocol/entity-specific regulator enforcement, court cases, and sanctions designation: Not verifiable as of October 10, 2026. Data protection. No protocol-specific privacy notice or identified data controller was found. On-chain addresses and transaction history are public and persistent; any web-interface telemetry or wallet/IP processing depends on the frontend and infrastructure provider. Not verifiable as of October 10, 2026. Risk conclusion: Legal exposure is concentrated in pseudonymous DAO governance, identifiable developers/service providers, frontend operators, oracle/liquidation administration, and users—not eliminated by the “decentralized” label.

GLEIF LEI registry unavailable at scan time. OFAC SDN screening of 'Swiss Stake AG', 'Curve LlamaLend': no match. SEC litigation and administrative release feeds: no mention.

Curve LlamaLend on Ethereum uses crvUSD as the protocol’s native stablecoin, so the question of depeg applies to crvUSD rather than a third-party asset. The gathered web sources do not provide a verifiable historical depeg series for crvUSD or for the specific LlamaLend markets, so depeg_count, last_depeg_date, and max_depeg_pct are not verifiable as of 2026-10-10. One Curve source explicitly states crvUSD held its peg within about $0.997–$1.000 through the cited market decline and had no depeg episode in that period, which supports stability for that window only, not full-history verification.

Curve LlamaLend’s principal risks are oracle quality, liquidation-driven bad debt, permissionless market configuration, incomplete independent assurance for v2, and administrative or wind-down risk. Ethereum-specific TVL, lender concentration, utilization, and current exposure are Not verifiable as of October 10, 2026 because Dune was unavailable; therefore, the matrix emphasizes structural and documented failure modes rather than current size.

Curve LlamaLend’s top strengths are: 1) Soft-liquidation protection via LLAMMA, which converts collateral gradually instead of forcing a single-price cliff liquidation, giving borrowers more time to react; 2) High capital efficiency, with very high LTVs reported for certain pairs and a design optimized for looping strategies; 3) Isolated, one-way markets, so risk, debt, oracle logic, and parameters are segregated per market and contamination risk is contained; 4) Permissionless market creation / flexible asset pairs, especially in v2, where markets are not limited to crvUSD-only pairs and can support broader ERC-20 combinations; 5) Composable infrastructure, including standard ERC-4626 vault integration and protocol-level risk controls such as configurable oracles, pausing, and safer vault accounting in v2.