Skip to content
PROTOCOL RESEARCH

Euler risk

Excellent core contracts and a fully repaid $197M V1 exploit, but V2 lets any vault deployer pick its own oracle, and that surface already blew up in November 2025 while TVL fell 83% from its peak.
CautionResearched Aug 5, 2026

Scorecard

strong: Audits & contractsSpearbit, OpenZeppelin, ChainSecurity, Certora, Omniscia, Trail of Bits, yAudit, Electisec, MixBytes, Zellic and more across EVC/EVK/oracle, plus a $1.25M Cantina competition with no high or medium findings and a $3.5M live-funds CTF that held 13 days.
watch: Admin controlEVC has no owner or admin on-chain, and the vault factory's upgrade authority sits behind a 4-day timelock, but each individual vault's governor is whoever deployed it.
weak: OracleVault deployers choose the oracle route; in November 2025 curated vaults carried xUSD near $1 against a market that fell to roughly a quarter of a dollar and then into single-digit cents, liquidations stalled and lenders absorbed bad debt.
watch: Liquidity & exitDeposits are down about 83% from the October 2025 peak, bad-debt socialisation is on by default, and exit is fine in blue-chip vaults but was not in the curated ones.
watch: Yield (real vs emission)Base supply APY is borrower-paid and real, but the 15 pools carrying reward APY hold roughly 76% of the TVL in DeFiLlama's Euler yield feed.
watch: Holder concentrationTop five pools are 72.7% of that feed's TVL and two K3 Capital Monad vaults alone are 58%; the feed covers only 44% of protocol TVL and EUL token distribution is _refresh_.
watch: Track recordV1 lost $197M in March 2023 and users were made whole at $240M; the V2 core has never been exploited, but its curated vaults took nine-figure bad debt in November 2025.
🟢 strong🟡 watch / caveat🔴 weak / fund-loss risk
Verdict is a gate (worst flaw wins), not an average. Our read, not financial advice.

Live facts

auto-sourced now
TVL$351.3M
30d↑8%
Audits2
Last hacknone

DeFiLlama + our exploits feed. Cross-check the dated report against today.

TL;DR

Euler V1 was drained for about $197M in March 2023 and the attacker gave it back. DefiLlama's hack registry records $197,000,000 taken and $240,000,000 returned against an entry named "Euler V1", and that protocol is wound down at $0. Today's V2 is a different codebase with an audit record few lending protocols match. The core has held. The layer above it has not. V2 is permissionless, so anyone can deploy a vault and choose its oracle route, and in November 2025 curated vaults valued Stream Finance's xUSD near a dollar while the market repriced it to roughly a quarter of a dollar inside 24 hours and into single-digit cents within days. Liquidations could not clear and lenders ate the hole. TVL peaked at $2.20B on 2025-10-06 and is $364M today.

Checklist

Audits & contracts. Reviewed to unusual depth: Spearbit, OpenZeppelin, ChainSecurity, Certora, Omniscia, Trail of Bits, yAudit, Electisec, MixBytes, Zellic, Ottersec, BailSec, CD Security, Supremacy and Hunter Security across the EVC, the vault kit, the price oracle and the periphery, with per-chain bytecode verification published in euler-interfaces. A $1.25M Cantina competition drew more than 600 participants and produced no high or medium findings; a $3.5M Hats capture-the-flag left live funds in Euler markets for 13 days without a loss, and a $500K EulerSwap live CTF in June 2025 ended with the posted mainnet liquidity untouched.

Admin control. The EVC has no owner and no admin; both calls revert on-chain. The vault factory does hold an upgradeAdmin that can swap the implementation for upgradeable proxies, and it resolves to Euler's own eVaultFactoryGovernor, whose single DEFAULT_ADMIN_ROLE holder is a timelock with a 4-day minimum delay. Protocol fee config sits behind a 4-of-8 Safe. Per-vault governors, however, are third parties chosen at deployment.

Oracle. Each vault's deployer picks the oracle route. Euler's own adapters are immutable and ungoverned, and a router can be finalized by revoking its governor, but it can equally stay governed, and a curator may point it at a hardcoded or exchange-rate source. This is the same delegated-oracle shape we rated red for Hyperliquid HIP-3 on 2026-08-05, with one difference that cuts against Euler: here it has already failed with real money on it.

Liquidity & exit. Deposits are down roughly 83% from the October 2025 peak. Bad-debt socialisation is enabled by default, so a shortfall in your vault is spread across its depositors rather than left with the borrower; it fires only once a liquidated account's collateral is exhausted, and a vault governor can switch it off, which makes it one more per-vault variable rather than a protocol constant. A mainstream USDC or WETH vault redeems normally. A vault whose collateral has stopped trading does not.

Yield: real vs emission. Base supply APY is paid by borrowers and is genuine. But the 15 pools carrying reward APY hold about $121M of the $159.5M in DeFiLlama's Euler yield feed, so roughly three-quarters of tracked deposits sit somewhere that is being paid to sit there.

Holder concentration. The top five pools are 72.7% of that feed's TVL and the top ten are 86.2%. Two K3 Capital Earn vaults on Monad hold $92.5M between them, about 58% of the tracked total. Read that with its limit in mind: the yields feed accounts for $159.5M, about 44% of the protocol's $364M, so this measures concentration among tracked pools rather than across all of Euler. EUL token distribution: refresh.

Track record. One catastrophic V1 exploit, repaid in full. No exploit of the V2 core since its 2024 launch. One nine-figure curator and oracle failure on V2 rails in November 2025 that the contracts survived and depositors did not.

Worst case

Two shapes, and they are not the same size.

The core failing is now unlikely but not impossible: a bug in the vault implementation reached through the factory's upgrade path. That requires the factory governor to act, which requires the 4-day timelock, so you would get warning.

The realistic one already happened. You deposit into an Euler-branded vault picked off a yield number, a stranger governs it, its oracle route holds a collateral asset at a price the market has abandoned, liquidations stall, and bad-debt socialisation hands you a slice of the shortfall. Nothing backstops this. The March 2023 repayment was one attacker's choice, not a protocol guarantee, and it does not repeat.

For our own board this is concrete, and it is no longer hypothetical. Euler shows 142 lending markets across 10 chains in DeFiLlama's borrow feed, and that feed carries no curator, governor or admin field at all. The yields feed's poolMeta does name a curator for Earn vaults ("K3 Capital Earn WETH") while for EVK vaults it is only an index ("EVK Vault eUSDC-70"). Our own pipeline then drops even that: the lending scanner stores protocol, chain and symbol and never persists poolMeta, and the board renders one protocol name per row. So a governed blue-chip vault and a stranger's permissionless vault already appear side by side on /lending under one name and one verdict, sorted by APY, with nothing in our data that could tell them apart. That is the November 2025 failure mode rebuilt in our own UI. Euler does publish labels and perspectives through euler-labels and its known-vault endpoints, so the join exists upstream; DefiLlama does not carry it and we do not fetch it. Until we do, our Euler rows are unattributed and should be labelled as such.

Bottom line

Caution, and the red is the oracle. The contracts are excellent and the V1 debt was paid in full, but Euler V2 is infrastructure rather than a product, and one brand now covers vaults with entirely different risk. A governed blue-chip vault on a major-provider feed is a reasonable place to sit. An unattributed high-APY vault is not the same instrument. Size by vault, not by protocol.

Data appendix

  • TVL: ~$364.16M for euler-v2 (orchestrator-verified via api.llama.fi/tvl; my independent chart pull agrees at $364.4M). Parent euler ~$364.18M. Euler V1 reads $25,890, effectively $0.0M, and is wound down, which is why the parent-child gap is only ~$26k. ATH $2,202.1M on 2025-10-06, down 83.5%. 90d prior (2026-05-08): $384.1M. Source: api.llama.fi/protocol/euler-v2.
  • Markets: 142 Euler markets in DeFiLlama's /lendBorrow feed across 10 chains on my pull (Ethereum 49, Base 21, Monad 16, Hyperliquid L1 10, BSC 10, Avalanche 9, Arbitrum 8, Linea 7, Plasma 7, Unichain 5); the orchestrator measured 141 earlier the same day, and the feed moves intraday. Separately, the yields feed carries 158 Euler pools across the same 10 chains (Ethereum 53, Base 22, Monad 21, Hyperliquid L1 13, Avalanche 10, BSC 10, Arbitrum 8, Linea 8, Plasma 8, Unichain 5). The two counts are different populations and should not be merged. Note that /lendBorrow itself carries no project or chain field; both figures require a join on pool UUID against /pools. The Ethereum vault factory has minted 874 proxies (getProxyListLength() = 0x36a, read on-chain).
  • Audits: EVC, EVK and price oracle reviewed across 2023-2026 by Spearbit, OpenZeppelin, ChainSecurity, Certora, Omniscia and Trail of Bits (EVC, 30 January 2024); further reviews by yAudit, Electisec, MixBytes, BailSec, CD Security, Supremacy and Hunter Security; FeeFlow by Zellic, Ottersec and Team Omega; fuzzing by the Victor Martinez suite; web-layer penetration tests by Ruptura and FailSafe. $1.25M Cantina competition: more than 600 participants, no high- or medium-severity findings, $200,000 allocated. $3.5M Hats CTF: no funds compromised over 13 days. Source: docs.euler.finance/security/audits.
  • Bug bounty: up to $7.5M via Cantina, scope covering EVK, EVC, the price oracle, EulerEarn, EulerSwap and the web interface, with a SEAL Safe Harbor path for live incidents. Source: docs.euler.finance/security/bug-bounty.
  • Admin/governance (read on-chain, Ethereum, 2026-08-05): EVC 0x0C9a3dd6b8F28529d72d7f9cE918D493519EE383 has no owner() or admin(), both revert. Vault factory 0x29a56a1b8214D9Cf7c5561811750D5cBDb45CC8e returns upgradeAdmin() = 0x2F13256E04022d6356d8CE8C53C7364e13DC1f3d, listed as eVaultFactoryGovernor in Euler's own registry; its only DEFAULT_ADMIN_ROLE member is timelock 0xfb034c1C6c7F42171b2d1Cb8486E0f43ED07A968 with getMinDelay() = 345600 seconds (4 days). protocolConfig 0x4cD6BF1D183264c02Be7748Cb5cd3A47d013351b returns admin() = 0xcAd001c30E96765aC90307669d578219D4fb1DCe, a Safe with getThreshold() = 4 and 8 owners. On-chain implementation() = 0x8Ff1C814719096b61aBf00Bb46EAd0c9A529Dd7D, matching eVaultImplementation in the published registry. Individual vault governors: deployment-specific, not enumerated this run (refresh).
  • Oracle: per-vault routes through EulerRouter. Euler's docs state adapters are "immutable, ungoverned, bidirectional contracts", while the router's "configuration can remain governed or be finalized by revoking its governor". Supports Chainlink, Chronicle, Pyth and RedStone plus onchain and rate-based sources, and can price ERC-4626 shares via convertToAssets. Euler's docs warn that "ERC-4626 does not require convertToAssets to resist manipulation". Source: docs.euler.finance/learn/price-oracles.
  • V1 incident: 2023-03-13, $197,000,000 taken and $240,000,000 returned, technique logged as "Flashloan Donate Function Logic Exploit" against an entry named "Euler V1". Source: api.llama.fi/hacks. Chainalysis independently confirms the vulnerable function was the eToken donateToReserve/donateToReserves path, that a liquidity check was missing, and that the funds were returned in full; the $240M figure reflects Euler's own accounting of total recovery value against $197M taken. Root cause per multiple post-mortems (Cyfrin, Hacken, QuillAudits): donateToReserves(), added in EIP-14, omitted the checkLiquidity() call, letting an attacker burn collateral without a solvency check while debt stood, force self-insolvency and self-liquidate into an oversized discount. Chainalysis does not itself name EIP-14. V2's EVK Vault module carries no such function (deposit, mint, withdraw, redeem, skim), and EVK tracks balances internally, which removes the donation-manipulation precondition for a vault's own shares. Euler's own docs state plainly that this "does not make every ERC-4626 collateral or oracle integration safe: external share vaults, oracle routes, recursive collateral paths, liquidity, and redeemability can introduce different risks". Source: docs.euler.finance/curate/vaults/evk/donation-attacks.
  • Bad-debt socialisation: Euler's docs state vaults "have bad debt socialisation enabled by default", that it "socializes losses to all depositors" to prevent bank-run dynamics, that it triggers only once the liquidated account's collateral is exhausted and a minimum liability threshold is met, and that "vault governors who do not want debt socialization can disable this functionality".
  • Vault types and labelling: escrow and finalized vaults have no active governor; governed vaults do. Classifications live in the euler-labels repo and are served through known-vault endpoints, carrying an explicit "Known does not mean safe" warning. Euler Labs states it "does not run or govern any governed EVK vault". Source: docs.euler.finance/learn/vault-types.
  • Holder concentration: top 5 pools 72.7% and top 10 86.2% of the $159.51M yields-feed TVL; two K3 Capital Monad Earn vaults total $92.5M (58%). The yields feed covers 43.8% of protocol TVL, so these shares describe tracked pools only. EUL token holder distribution: refresh (Etherscan top-holders is a Pro endpoint and returned NOTOK).
  • Recent-news scan: the November 2025 Stream Finance / xUSD collapse propagated into Euler curated vaults. Stream disclosed roughly $93M of losses and froze withdrawals on 2025-11-04; reporting is consistent that lending protocols held xUSD at a hardcoded $1.00, which stalled liquidations. xUSD fell about 77% to a 24-hour low near $0.26 and was trading around $0.07 to $0.14 with almost no liquidity by 2025-11-08; the precise low varies by venue and moment (refresh). Re7 Labs acknowledged roughly $14.65M exposure in its xUSD isolated vault on Euler. A widely repeated figure puts Euler-side bad debt near $137M and ecosystem-wide direct exposure near $285M; neither was confirmed against a primary source this run (refresh). Reporting also indicates the affected vaults were third-party curated rather than Euler-operated, consistent with Euler Labs' own statement that it governs no EVK vault.
  • TVL trajectory, primary and independently pulled: $2,202.1M on 2025-10-06, $1,612.0M on 2025-11-01, $1,068.4M on 2025-11-10, $1,029.1M on 2025-12-01, $384.1M on 2026-05-08, $364.4M today. The November window accounts for part of the fall, not all of it; the decline continued through 2026.
  • Bad debt since V2 launch: the November 2025 event is confirmed qualitatively by multiple independent reports. A current protocol-wide outstanding bad-debt total: refresh.

Maintained monthly. Methodology: DeFi Research Instruction v2.

← all protocols