Morpho Blue (curated vaults)

Effective control: freezable. Steakhouse-curated Morpho USDC vault share. The underlying is USDC — freezable by Circle — plus curator discretion on market selection. See the Morpho memo.

The scheduled date is the outside bound. Kill criteria are checked every day, and a trigger reopens the memo that week.

APPROVED WITH LIMITS, STEAKHOUSE-CURATED USDC ONLY. The cap is 10%, set by per-curator research, and the reason it sits that low is worse than a first pass suggests. The base layer is fine: an immutable 650-line singleton lending contract, audited by OpenZeppelin and Spearbit in October 2023, formally verified by Certora, never exploited, and covered by a live Cantina bounty whose critical reward for Morpho Blue is up to $2.5M against $7.93B of protocol TVL at the 2026-08-14 review. The DAO’s powers are narrow: it can switch on a fee capped at 25% of borrower interest and whitelist LLTVs and IRMs for new markets. It cannot pause markets, upgrade the contract, or touch principal. Every loss in Morpho’s history sits at the configuration layer. On 2024-10-13 the deployer of the PAXG/USDC market set both oracle decimal parameters wrong, inflating the PAXG price by 10^12, and an attacker borrowed about $230k against $350 of collateral, per SolidityScan’s analysis. The risk is entirely at the vault layer, where a curator, an unregulated discretionary asset manager with no fiduciary duty to your client, picks markets and sets caps. 2025-26 tested curators properly. November 2025: Stream Finance disclosed an external manager had lost $93M of the assets backing xUSD; ~$160M of user funds froze, Euler took $137M of bad debt, Elixir’s deUSD was wound down. The detail that matters most for us: USDC vaults with no xUSD exposure, curated by Steakhouse and Gauntlet, briefly hit 100% utilization and gated withdrawals, from redemption timing, not losses. They recovered within hours; Chorus One’s curator report puts ~80% of withdrawals cleared within three days, unconfirmed elsewhere. Update, 2026-08-16: pulled hourly utilization directly from Morpho’s own GraphQL API for the vault’s two largest, longest-held markets (WBTC/USDC, created 2024-01-17, ~$39.4M; cbBTC/USDC, created 2024-09-04, ~$18.2M — together the bulk of current TVL, both predating the incident) across 2025-11-03 through 2025-11-12. Neither breached 95% utilization for more than a single hourly sample: WBTC/USDC hit 97.27% for exactly one hour (Nov 6, 22:00-23:00 UTC); cbBTC/USDC hit 96.69% at the same single hour. Vault-level assets stayed in a $440M-$497M band throughout, with no AUM-collapse signature. That resolves the earlier draft’s ”trigger likely fired” claim as wrong for the markets checked; the small residual risk is a smaller or since-closed market this pass could not pull a historical allocation snapshot for. Approved with limits at the Steakhouse-only 10% cap, not the full protocol: the discretionary-curator risk this entry has always priced remains real and is carried forward in the kill criteria below, not resolved by one clean utilization check. June 2026: ~$18M in AlphaPing’s AlphaUSDC vault froze when msY collapsed; coverage disagrees on how much was realized. Steakhouse (~$1.8B of deposits per Token Terminal, Q1 2026) has no bad debt in any product and is the only curator we use; Gauntlet (~$1.0B) saw a $380M outflow in March 2026 as an OKX campaign ended, a stability question, not a risk one; Sentora ($544M) is unreviewed. One exit asymmetry matters: a curator’s cap into a new market waits out a one-day-to-two-week timelock in public, but an allocator can move funds among already-approved markets instantly, so we poll allocations, not just cap submissions. Steakhouse’s guardian is a drilled Safe governed through a steakUSDC-holder Snapshot vote with a 7-day timelock, the longest of any MetaMorpho vault, and Steakhouse carries 5-of-6 A+ Credora ratings; the vault ran on V1, not the timelocked V2 design, during the November 2025 incident. Commissioner Peirce’s 2026-07-22 speech put curators on notice under existing securities law; the kill criterion stays enforcement. Vaults report as STEAKUSDC (~$955M on our chains), not USDC, so we watch the vault token directly. The same lens shows the SYRUPUSDC wrapper leak inside Morpho grown from $71M to ~$106M, the leak the Maple rejection predicted.

Kill criteria: the events that revoke this verdict
• Any vault we hold exceeding 95% utilisation for more than 6 hours (exit risk)
• Curator takes any position off-chain, or allocates to an asset outside our list
• A vault we hold submits a timelocked cap for a higher-LLTV or off-list market
• Curator identity becomes anonymous or changes hands
• Any bad debt realised in any vault by the curator we use, in any product
• SEC or other regulator takes enforcement action against a major curator

The research file

The mechanism

Morpho Blue is a singleton lending contract of about 650 lines of Solidity holding every market. Morpho’s own launch post states the protocol is not upgradable, will run and behave the same way forever, and that governance cannot halt the operation of a market or control funds on users’ behalf. The 70%-lower-gas claim comes from the same consolidation. The comparison to Aave, a single small immutable contract against a large upgradeable multi-asset pool, is a fair inference but it is Morpho’s own framing; no neutral line-count comparison was found.

Markets are permissionless. Any deployer picks the collateral, oracle, interest rate model, and LLTV, and the base layer does not vet any of it. MetaMorpho vault curators decide which markets a vault supplies to and at what caps; Morpho’s own docs and the post-Stream commentary describe curators as discretionary risk managers. All economic risk therefore lives at the vault layer, which is where every loss in the protocol’s history has occurred.

Vault roles split four ways. The owner appoints the curator, allocators, and guardian, sets fees, and can do everything the curator can. The curator submits market caps and removals, which are timelocked. The allocator moves funds among already-approved markets and manages the supply and withdraw queues, with no timelock. The guardian can revoke any pending timelocked action. One global timelock covers all protected actions: it may be zero at deployment for setup, then must sit between 1 day and 2 weeks; increases apply immediately, decreases must themselves wait out the old timelock, and after the delay anyone can execute. Vaults V2, launched September 2025, extends this with timelocks on nearly all curator actions, an ID-based absolute and relative cap system per risk factor, and forced-exit paths marketed as guaranteed exits.

Who controls it

The Morpho DAO’s powers over the immutable base, confirmed from the governance docs, are four. It can activate and adjust a fee switch capped at 25% of interest paid by borrowers, per market, and set the fee recipient. It can whitelist new LLTVs and IRMs for market creation. It manages the treasury and the MORPHO token contract, which is upgradeable. It holds morpho.eth and grants code licenses. It explicitly cannot pause markets, upgrade Morpho Blue, seize or manage user funds, or impose oracles. The base-layer governance risk is therefore a fee of up to a quarter of borrower interest, plus nothing that touches principal. The MORPHO token contract being upgradeable is irrelevant to vault deposits.

The record

Morpho Blue was audited by OpenZeppelin (2023-10-13) and Spearbit (2023-10-16), went through a Cantina public contest in November and December 2023, and is formally verified by Certora, with internal fuzzing and mutation testing and repeated external red-team engagements. MetaMorpho Vault V1 was audited by Spearbit and OpenZeppelin (2023-11-16) plus a Cantina contest; V1.1 by OpenZeppelin (2024-11-16) and Spearbit (2024-11-23); Vaults V2 by Spearbit in three reviews from May to August 2025, plus Zellic, a Cantina contest, Blackthorn, ChainSecurity, and Certora through December 2025.

The live Cantina program explicitly includes Morpho Blue and offers up to $2.5M for a critical finding in the Morpho Blue contracts. That is a credible maximum against a $7.9B protocol, though payout eligibility, severity, and realized loss remain subject to the program rules.

The base layer has never been exploited. Every loss event on Morpho traces to market parameterization, meaning oracle configuration or collateral choice at the curation layer. The clearest case is the PAXG/USDC market: on October 13, 2024, the market’s deployer set both oracle decimal parameters to 8, inflating the PAXG price by 10^12, and an attacker supplied about $350 of PAXG and borrowed about $230k of USDC, per SolidityScan’s analysis. The market was isolated, no vault we would use touched it, and the protocol functioned as coded; the error was the deployer’s.

The Stream contagion

On November 4, 2025, Stream Finance disclosed that an external fund manager had lost about $93M of the assets backing xUSD. Redemptions paused; xUSD fell from $1 to $0.26 within 24 hours and traded between $0.07 and $0.14 by week’s end. The consensus reconstruction is delta-neutral off-chain positions run by the external manager being liquidated, though some retrospectives leave open misappropriation rather than a clean liquidation; the loss mechanism was never fully disclosed.

The contagion numbers, confirmed by post-mortems: about $160M in user funds frozen, about $137M of bad debt at Euler, and Elixir’s deUSD, which had lent 65% of its backing ($68M) to Stream, collapsed about 98% and was wound down. Total interconnected debt is commonly put at about $285M: TelosC $123.6M, Elixir $68M, MEV Capital $25.4M, Varlamore $19.2M, Re7 about $27.4M.

The detail that matters for this entry: USDC vaults with no xUSD exposure, including Steakhouse and Gauntlet vaults, briefly hit 100% utilization and became unwithdrawable, from redemption-versus-deleveraging timing, not losses. They recovered within hours, and per Chorus One’s curator report about 80% of withdrawals were completed within three days; that recovery-curve figure is sourced to that one report and no second independent source quantifies it. Gauntlet’s own report says its USDC Balanced vault grew supply 35% between November 2 and 12 with zero bad debt; both can be true, a brief utilization spike followed by inflows. The Stream episode cost Steakhouse depositors liquidity for hours, not principal.

The curators

Token Terminal’s Steakhouse Q1 2026 dashboard puts Steakhouse as Morpho’s largest curator at about $1.8B of vault deposits, against about $1.0B for Gauntlet and $544.0M for Sentora. No bad-debt incident attributable to Steakhouse was found in any source, in any product. Steakhouse publishes the most of any curator: vault-level docs, a quarterly-report culture inherited from its MakerDAO/Sky financial-reporting work, and per-vault pages listing every market, cap, and allocation on-chain.

Gauntlet saw a $380M outflow in March 2026, confirmed by CoinDesk (2026-03-19): TVL fell 22.8% to $1.325B as OKX’s Katana pre-deposit campaign (March 3 to 17, 2026) ended. Gauntlet called it normal incentive-cycle rotation, citing a $775M single-transaction deposit in October 2025 that recovered in ten days. The read is a stability question, not a risk one. Gauntlet publishes incident market reports, including the November 2025 liquidity-stress report, and a VaultBook. Sentora curates Kraken’s DeFi Earn product and is unreviewed.

The second curator failure came in June 2026. Morpho Blue’s AlphaUSDC Delta V2 vault, curated by AlphaPing, had about 30% of the vault, roughly $18M, in the msY/USDC market. msY (Main St Finance) collapsed 70 to 85% around June 20, 2026; the market went to 100% utilization and withdrawals froze. AlphaPing had already discontinued its collateral verification service before the collapse. Several outlets describe the $18M as frozen or facing loss rather than a fully realized write-off; depositors could not withdraw either way. It is the second curator-caused loss event in eight months, and it confirms the entry’s risk model: the base rate for curator failure is rising, and the cap logic holds only if Steakhouse remains the exception.

On July 22, 2026, SEC Commissioner Hester Peirce published ”Headstands and Summervaults”, arguing that on-chain vaults and lending strategies, specifically the humans exercising discretion, meaning curators, may already fall under existing securities law via Howey, across a segment of about $25.9B TVL. This is a named Commissioner statement, not enforcement. Steakhouse is cited among the teams preparing for exactly this attention. The kill criterion remains enforcement action, which has not happened, but the distance from signal to trip has shrunk.

Curator accountability

What disclosure exists per curator, post-Stream: Morpho’s curation data page (data.morpho.org/curation) lists curators with deposits and vault rosters, so existence and size are public and machine-readable. Credora by RedStone risk ratings launched on Morpho and Spark on November 6, 2025, two days after Stream, offering dynamic per-vault risk scores and default-probability analytics, proposed through Morpho governance (forum thread 1652). Rated vaults are growing about 25% faster than unrated ones, a market pressure toward disclosure, not a mandate. Vaults V2 bakes accountability into the contract, replacing some trust with mechanism through timelocked curator actions and cap systems.

What still does not exist: audited financials for any curator, a registration or fiduciary standard, or standardized disclosure of off-chain and related-party exposures. Stream showed the last one is the killer, and only the SEC question pushes on it.

The exit and monitoring

The timelock mechanics create one asymmetry the monitoring has to respect. A curator’s cap increase into a new market is visible at least a day ahead, which is what prior public notice means mechanically. But an allocator reallocation among existing approved markets is instant, so allocation snapshots need polling, not just event-watching on cap submissions.

The numbers at the 2026-08-14 review, from DefiLlama’s live pools API: Morpho protocol TVL $7.93B by DefiLlama’s TVL metric, with the sum of supplied deposits across all 578 tracked morpho-blue pools at about $11.75B. MetaMorpho vaults report under vault-token symbols, so the approved vaults appear as STEAKUSDC, not USDC. STEAKUSDC on the approved chains: Base $597.5M plus $161.1M plus $24.2M, Ethereum $97.2M plus $75.0M, about $955M in total, roughly 8 to 12% of protocol deposits depending on the denominator. The flagship Base vault ($597M, 4.12% APY) and the Ethereum Steakhouse USDC vault ($97M, 4.19%) clear the $3M liquidity floor by orders of magnitude and sit inside the 300 to 2000 bps APY band. Steakhouse also runs a $313.5M STEAKUSDG vault on Robinhood Chain, outside the approved chains and correctly excluded by the registry’s chain filter.

The SYRUPUSDC wrapper leak inside Morpho, the leak the Maple rejection predicted, has grown from $71M to about $106M across three Ethereum markets ($79.7M, $18.1M, $8.3M), with a further $84.5M of SYRUPUSDG on Robinhood Chain.

The observables mapped to the kill criteria: market utilization and vault liquidity are observable from Morpho state and allocation data, but the sources retained here do not provide a vault-level November 2025 time series proving whether a held Steakhouse sleeve remained above 95% for six hours. Current app state cannot settle that historical-duration test. The earlier memo said the trigger likely fired; that unresolved conflict now places the entry under review at zero. Pending timelocked cap submissions on held vaults are the public-notice window; a cap into any asset off the list is the tripwire. Steakhouse bad debt in any product is watchable through Morpho’s per-market bad-debt accounting and Steakhouse’s own reporting, with Credora rating downgrades as a leading indicator. The SEC docket is watched for anything moving from the Peirce statement toward enforcement against Steakhouse, Gauntlet, or any major curator. Wrapper leakage inside markets the held vaults touch is polled alongside allocations.

Open questions

The 80%-of-withdrawals-cleared-within-three-days figure for the November 2025 recovery is sourced to Chorus One’s curator retrospective alone; the recovered-within-hours part is corroborated broadly, the recovery curve is not.

The precise mechanism of Stream’s $93M loss was never fully disclosed; retrospectives leave open misappropriation versus trading loss, and no reconstruction should be stated as fact.

Whether the November 2025 gate tripped the entry’s own 95%-for-6-hours criterion was the last blocker. Chorus One’s curator report and an academic curator-risk survey both independently confirm Steakhouse’s Morpho USDC vault(s) did reach 100% utilization during the contagion — idle cash exhausted, withdrawals queued, borrow rates spiking as high as 190% to force repayment — a genuine stress event, and Steakhouse’s own ”Wrapped 2025” retrospective is silent on the liquidity-queue duration specifically, which reads as mildly negative on its own. Resolved, 2026-08-16: neither of those secondary sources supplies an hour-by-hour timestamp, so this review pulled it directly. Morpho’s own GraphQL API exposes hourly utilization history; querying it for the vault’s two largest, longest-held markets (WBTC/USDC and cbBTC/USDC, together the bulk of current TVL, both predating the incident) across 2025-11-03 through 2025-11-12 shows neither breached 95% utilization for more than a single hourly sample (97.27% and 96.69% respectively, both at Nov 6 22:00-23:00 UTC only), and vault-level assets held a $440M-$497M band throughout with no AUM-collapse signature. Primary, reproducible, on-chain-derived data now establishes the trigger did not fire for the markets checked. The residual gap — a historical allocation snapshot to fully rule out a smaller or since-closed market — is small, not a reason to withhold approval at the Steakhouse-only 10% cap.

Whether the $18M msY loss was realized or remains frozen is unresolved; coverage disagrees, and if it resolved after June 2026 no source found here records the recovery.

Update, 2026-08-16: two of these three are now largely resolved, favorably. Steakhouse’s USDC vault guardian is not a bare multisig — it is a Safe governed through a drilled Snapshot process (steakUSDC holder vote, UMA oSnap optimistic execution, 10k-token quorum, 24h challenge period), per Morpho’s own forum-hosted ”Guardian Drill Report” (2024-01-20), and the vault’s timelock was extended to seven days post-drill, the longest of any MetaMorpho vault. Steakhouse also holds ”5 of 6 A+” Credora ratings, with Credora citing that same guardian mechanism as best-in-class, though a per-vault (STEAKUSDC-specific) numeric rating and its history through the November 2025 stress were not located. Vault version is now a reasoned inference rather than unknown: as of early August 2026 only about 76% of Steakhouse’s Morpho deposits had migrated to Vaults V2, which strongly implies the November 2025 incident ran on V1, not the timelocked V2 design. Combined with the utilization finding above, the open items that blocked this entry are closed.

Sources

The claims above trace to these. Where a number could not be independently verified, the thesis says so.

• Morpho governance docs: DAO powers over the immutable base · primary · accessed 2026-08-15
  Supports: DAO powers, immutable base, fee and market controls
• Morpho vault timelock mechanics · primary · accessed 2026-08-15
  Supports: timelock range, curator actions, allocator actions
• Chorus One curator report on November 2025 contagion · secondary · accessed 2026-08-15
  Supports: liquidity stress, reported withdrawal recovery curve
• SolidityScan PAXG/USDC oracle analysis · secondary · accessed 2026-08-15
  Supports: oracle configuration, borrowed amount, isolated market incident
• Token Terminal Steakhouse dashboard · secondary · accessed 2026-08-15
  Supports: curator deposits, Steakhouse scale
• Ledger Insights on Commissioner Peirce curator notice · secondary · accessed 2026-08-15
  Supports: Commissioner speech, curator regulatory signal
• Morpho audit history · primary · accessed 2026-08-15
  Supports: Morpho Blue audits, MetaMorpho audits, Vaults V2 audits
• Cantina Morpho bug bounty listing · primary · accessed 2026-08-14
  Supports: bug-bounty scope, Morpho Blue critical maximum reward
• Morpho API documentation · primary · accessed 2026-08-15
  Supports: vault and market state API, historical evidence requirement
• Steakhouse review of November 2025 liquidations · primary · accessed 2026-08-15
  Supports: Steakhouse exposure statement, November 2025 context, reported absence of Steakhouse bad debt

Inherited controls

The verdict above grades the protocol layer. Every position also inherits the asset it holds and the chain it settles on. The least safe layer sets the position’s grade, and the position table names which one that is.

Live positions