CVE Index中文
L4 · Septic Tank BlewKava Lend · DeFi · Lending poolskava-lend-busd-pool

Kava Lend exploit drains most of the protocol’s KAVA lending pool

Kava says the incident was confined to Kava Lend. Bitquery traced 55.2 million KAVA borrowed through repeated loans against inflated BUSD collateral and estimated about $2.1 million in sale proceeds on BNB Chain.

Disclosed
Nov 14, 2026
Updated Nov 17, 2026
Severity
Septic Tank Blew
No CVE: a breach, not a bug
Splash zone
Kava Lend users, especially KAVA lenders, and users whose transfers depend on paused exchanges or the Axelar bridge.
crypto
In the wild
Yes
Kava confirmed the incident. Bitquery traced the loans and ensuing sales.

What happened

Bitquery found that the attackers put a nearly empty BUSD market in a state where its interest calculation divided by a tiny positive supply. The supply index grew about 243,000-fold in three minutes. One wallet recycled the available BUSD into a second wallet’s account as collateral. The second wallet borrowed KAVA against those deposits. Repeated liquidations closed the debts after the first wallet had withdrawn the available BUSD, leaving little or none for the liquidation to seize. This is Bitquery’s reconstruction. Kava subsequently confirmed the accounting vulnerability and said the exploit was blocked.

Bitquery counts 55.2 million KAVA borrowed across 127 loans and 53.9 million KAVA retained after repaying other users’ loans, worth approximately $3.7 million at the pre-sale price. Its estimate of approximately $2.1 million measures the proceeds from sales on BNB Chain. The 612,416 USDT order that Relay blocked is part of the traced proceeds. Kava subsequently confirmed the amount taken. Final restitution is undecided.

Kava treated the Axelar suspension as a precaution. The assets in the sale reconstruction crossed through LayerZero. A bridge pause alone does not identify the exploit path or the status of customer balances.

Kava confirms about 49.2 million KAVA was bridged to BNB Chain through LayerZero. The underlying chain continued operating and the exploit created no KAVA. The amount frozen and any restitution remain unconfirmed. The planned permanent halt to new lending and borrowing requires a seven-day governance vote. On-chain proposal 223 covers the narrower step of freezing borrowing across current markets and restoring the USD 10 minimum-borrow parameter. Voting is open. The payload does not disable new deposits or settle restitution. Kava says it will issue separate instructions for remaining positions. Follow its official governance and recovery notices.

Crime-scene timeline

  1. Nov 14, 04:04–05:42 UTC

    Bitquery traces repeated borrowing and liquidation

    Two wallets used the BUSD pool to inflate collateral and borrowed about 55.2 million KAVA across 127 loans.

  2. Nov 14, 2026

    Kava confirms the incident is confined to Kava Lend

    Kava recognized the bridging and sale of KAVA on BNB Chain and ruled out unlimited minting. Partner exchanges paused deposits and withdrawals.

  3. Nov 14, 2026

    Kava pauses Axelar withdrawals as a precaution

    Kava paused the bridge and said a full incident report would follow. The traced sale route was through LayerZero.

  4. Nov 14, 16:20 UTC

    Bitquery observes a failed Relay order

    Relay showed a blocked order for 612,416 USDT with no payout. The record showed a possible return to the sender. Recovery by Kava or users was unverified.

  5. Nov 16, 2026

    Kava confirms amount taken and planned governance vote

    Kava confirms about 53.9 million KAVA was taken. The exploit is blocked, some funds were frozen, and permanent retirement of new lending and borrowing requires a governance vote.

  6. Nov 16, 18:29 UTC

    Governance proposal 223 enters voting

    It would set all current Kava Lend borrowing caps to zero and restore the USD 10 minimum-borrow parameter so existing borrowers can make partial repayments under the normal minimum-debt rule.

Who said what

We list each claim and who made it. Primary sources (the vendor, the researcher, the registry) are marked. Where reports disagree, we grade by the worst credible case.