Moonwell drained for about $8.7 million in a MAMO price-manipulation attack — its fourth pricing failure in under a year

Moonwell, a lending protocol on Coinbase's Base network, said on Thursday it was investigating an issue affecting its MAMO Core Market after blockchain security firms flagged an attack that moved millions of dollars of user deposits out of the protocol.

The mechanics were not exotic. CertiK said the attacker manipulated the collateral price of MAMO — a relatively illiquid token accepted as collateral on Moonwell — and then borrowed real assets against the inflated position. CertiK specifically identified borrowing of cbBTC from the mCBTC market. Protos reported a wider set of assets walked out the door: cbBTC, USDC, WETH and wstETH. Blockaid independently identified the same attack mechanism, according to The Block, and Protos said Blockaid flagged the attack roughly an hour after it began, initially sizing the damage at "$4+ million" before the total climbed.

Two firms converge on the final number. PeckShield and CertiK both put losses at approximately $8.7 million, per The Block. PeckShield said the attacker consolidated the proceeds into DAI at a single address. Protos, whose headline rounds the figure to $9 million, described the same endpoint: 8.7 million DAI — a stablecoin whose issuer cannot freeze balances — sitting in an Ethereum address funded through Tornado Cash. That the DAI figure and the loss estimate match at 8.7 million is the strongest corroboration available; the attacker's on-chain balance and the security firms' independent accounting agree.

The response

Moonwell's public response, posted to X and quoted by both outlets, was a blunt circuit breaker: "As a precaution, borrow caps for all Core Markets on Base have been set to 1 wei, preventing new borrowing and limiting the potential for further impact." Supply caps for MAMO and for WELL, Moonwell's own governance token, were also cut to 1 wei. All other supply caps were left unchanged.

Setting a cap to 1 wei is the smallest non-zero value the contracts accept — functionally a halt. Applying it to all Core Markets on Base, not just MAMO, means the team either could not immediately scope the blast radius or chose not to bet on its own scoping. Moonwell said it would share further updates when it had more information. The protocol did not respond to The Block's request for comment.

What the attack cost the attacker

Protos reported the attacker spent around $7 million pumping MAMO's price and sold the position afterward for an estimated $3.8 million loss. Set against $8.7 million extracted, those figures imply a net take in the neighbourhood of $4.9 million — an arithmetic inference from Protos's numbers, not a figure either outlet published, and neither the pump cost nor the exit loss has been independently confirmed by a second firm. The Block did not report the attacker's cost basis at all.

Token markets reacted, though not dramatically. WELL was down roughly 13% over 24 hours per CoinGecko, and MAMO down roughly 9% per DEX Screener, both as cited by The Block.

The pattern

This is the part that matters more than the dollar figure. Protos counted this as Moonwell's fourth incident in under a year, and noted all four were pricing-related:

• October 10: a price discrepancy during the market crash led to $12 million in liquidations and $1.7 million of bad debt.
• November: fallout from the Balancer hack broke the wrsETH/ETH oracle, leaving $3.7 million of bad debt.
• February: a Moonwell contract that Protos characterised as "vibe-coded" valued cbETH at $1.12 — reading the figure 1.12 as dollars rather than as a multiple of ETH — when the true value was around $2,200. Positions collateralised with cbETH were liquidated abruptly, adding $1.8 million of bad debt.
• Today: approximately $8.7 million drained via MAMO collateral price manipulation.

Those three prior bad-debt figures sum to $7.2 million; add Thursday's loss and the running total of value destroyed or extracted through Moonwell's pricing layer over roughly ten months is about $15.9 million. That sum is arithmetic on Protos's reported figures, not a number Moonwell has published or acknowledged.

Key facts

• Losses estimated at approximately $8.7 million — CertiK and PeckShield, via The Block; Protos headlined "$9M."
• Attack method: manipulation of the illiquid MAMO token's collateral price, then borrowing against it — CertiK; corroborated by Blockaid, via The Block and Protos.
• Assets borrowed: cbBTC (CertiK, via The Block); cbBTC, USDC, WETH and wstETH (Protos).
• Proceeds swapped to DAI and held at a single Ethereum address funded via Tornado Cash — PeckShield via The Block; Protos.
• Blockaid flagged the attack about an hour after it started; its first estimate was over $4 million — Protos.
• Attacker reportedly spent ~$7 million pumping MAMO and exited that position at an estimated $3.8 million loss — Protos.
• Borrow caps for all Base Core Markets, plus MAMO and WELL supply caps, set to 1 wei — Moonwell on X, quoted by The Block and Protos.
• WELL down ~13% and MAMO down ~9% over 24 hours — CoinGecko and DEX Screener, via The Block.
• Fourth pricing-related incident in under a year, following October ($1.7M bad debt), November ($3.7M) and February ($1.8M) — Protos.

The real-world read

"Investigating an issue" is doing a lot of work. Moonwell's own statement describes an "issue affecting the MAMO Core Market" and frames the caps as a "precaution." Two security firms and a third monitoring service describe an attacker manipulating a price and borrowing against it. Those are not the same register. The protocol has not, as of publication, used the words exploit or attack in its public statement.

The listing decision is the untouched question. Every account agrees MAMO is "relatively illiquid." Accepting a thin-float token as collateral in a market that also holds real cbBTC, WETH and wstETH is a governance choice someone made, through some process, with some risk parameters attached. Neither outlet reports who proposed it, how the price feed was constructed, or what the collateral factor was. Moonwell hasn't said.

Four for four on pricing. A single oracle failure is a bug. Four pricing-related incidents in ten months — a crash-time discrepancy, a contagion from someone else's hack, a unit-conversion error in a contract Protos called "vibe-coded," and now a manipulable collateral asset — is a pattern in how this protocol values things. The February incident is the most damning detail available, because a contract that confuses $1.12 with 1.12 ETH is not a sophisticated adversary problem; it is a review problem.

Nobody has said who eats it. There is no statement on whether suppliers are made whole, whether a treasury or insurance backstop exists, whether an on-chain post-mortem is coming, or whether the funds are recoverable. DAI in a Tornado-funded address suggests the practical answer on recovery, but that is inference, not confirmation.

Discount the AI framing. The Block wrote that the current run of DeFi exploits — more than $600 million across multiple protocols since April, led by the $292 million Kelp DAO hack — is "seemingly driven by advancements in artificial intelligence." That is an unsourced causal claim in an otherwise carefully attributed report. The attack described here required capital and a thin order book, not a model.

Read the disclosures. The Block discloses that Foresight Ventures is its majority investor and that Bitget is an anchor LP for Foresight; the outlet states it operates independently. Neither piece here is sponsored or a press release, and neither reads promotionally — but the disclosure is on the page and belongs in front of the reader.

Opinion, and whose

• CertiK and PeckShield (security vendors, both) each put the loss at ~$8.7 million and attribute it to MAMO collateral price manipulation. Estimates, arrived at independently, and they agree.
• Blockaid identified the same mechanism and first sized the loss at over $4 million — an early figure that the firm's own later reporting superseded.
• Protos characterises the February contract as "vibe-coded" and frames today's loss as the latest in a run of four. That framing is Protos's editorial judgment, though the underlying incidents are its own prior reporting.
• The Block advances the AI-driven-exploit-wave thesis without a named source for it.
• Moonwell has offered no assessment of cause, no loss figure of its own, and no remediation plan — only the caps and a promise of updates.

Sources

• The Block (James Hunt, 27 August 2026) — Moonwell's X statement and the 1-wei cap details; CertiK and PeckShield $8.7 million estimates; the cbBTC/mCBTC borrowing path; Blockaid corroboration; WELL and MAMO price moves per CoinGecko and DEX Screener; the $600 million/Kelp DAO sector context and the AI attribution. Carries a standing disclosure that Foresight Ventures is The Block's majority investor. Not sponsored.
• Protos (27 August 2026) — the ~$9 million headline figure; the Blockaid alert timing and its initial $4M+ estimate; the full borrowed-asset list; the $7 million pump cost and $3.8 million exit loss; the DAI/Tornado Cash destination; and the timeline of the October, November and February incidents, drawn from its own earlier coverage. Not sponsored; the page carries generic display advertising.
• Moonwell, via its post on X, quoted by both outlets — the only primary statement from the protocol available at publication.

No press release, sponsored post or commissioned research was used in this report.

Nothing here is financial advice.