# Kava shuts down HARD Lend after losing 53.9 million KAVA

> The exploit of an inactive BUSD market left 53.9 million KAVA exposed, and Kava is preparing a vote to shut down HARD Lend for good.
> By José Maldonado · DeFi · 2026-11-17

A lending market can look safe because it is empty. HARD Lend, the credit protocol on the Kava network, had a BUSD market that almost nobody used once that stablecoin lost its usefulness, and that abandonment was the entry point. On November 14, 2026, an attacker inflated his collateral roughly 243,000 times and drained the KAVA inventory; on November 16, Kava published the official update on the incident and announced it is accelerating the protocol's closure.

The figure that sums up the damage is 53.9 million KAVA, the net balance left in the attacker's hands after repaying third-party loans. Before the attack, those tokens were worth approximately 3.76 million dollars, with KAVA around 0.068 dollars. The token's price fell between 40 and 50 percent in the following hours, and that move explains much of the media impact.

The debate now is about governance. The network wants a vote on [Kava HARD Lend](https://cryptobriefing.com/kava-hard-lend-wind-down-exploit/) to permanently disable new lending and borrowing on HARD Lend, which will turn the protocol into an entity in liquidation rather than a live product. It is an unusual decision in DeFi, where teams usually patch the contract and carry on as if nothing had happened.

## How do you inflate collateral 243,000 times?

The explanation for the flaw is accounting, not cryptography. The attacker began by repaying all outstanding loans in the BUSD market until he was the only participant left. With the market in his hands, he left the net supply at a tiny amount, on the order of two hundred-millionths of a dollar, and there the protocol's interest calculation did the rest.

With such a small supply in the formula, the attacker's position multiplied roughly 243,000 times, according to [Immunefi's reconstruction for crypto.news](https://crypto.news/kava-lend-exploit-busd-market-immunefi/). That inflated collateral let him borrow all the available KAVA, and he did so in 127 lending rounds spread over 98 minutes, according to [Bitquery's reconstruction](https://bitquery.io/investigations/kava-hack) with two wallets working together.

The liquidation mechanism closed the circle. One wallet withdrew the underlying BUSD before the loan was liquidated, and the liquidation found almost nothing to seize. When the debt was canceled, the attacker regained borrowing capacity and repeated the cycle. Each round pulled KAVA out of the pool without adding new capital.

The tokens left the ecosystem through the usual route. About 49.2 million KAVA were bridged to BNB Chain with LayerZero, and some 48.4 million were sold there for 2.09 million dollars. The gap between book value and the cash obtained explains why the token's price collapsed so quickly.

The gross loan in the attack was 55.2 million, and the net came to 53.9 million after repaying what belonged to other users. That distinction between gross and net matters for sizing the protocol's real hole.

## The empty market as a vulnerability

Auditors look at markets with volume because that is where the money is. The HARD Lend case shows the reverse: a market with no activity can retain the accounting logic designed for a full market, and that mix produces absurd results. If total supply tends toward zero, any per-unit interest formula spirals.

Gonçalo Magalhães, head of security at Immunefi, noted that the attacker prepared the ground for three days with small deposits and withdrawals in that market. That public on-chain rehearsal went unnoticed, and the actual attack lasted less than two hours. The testing phase is the part any monitoring system could have detected if it had looked at the right market.

The case recalls other accounting errors that have defined the year. [Arcus lost 2.23 million dollars due to an accounting error](https://observatorioblockchain.com/en/defi/arcus-error-contable-2-23-millones-robinhood-chain/) a few days ago, and [Ankr's 303,000-dollar loss from reusing a deposit](https://observatorioblockchain.com/en/defi/ankr-puente-ethereum-303000-deposito-optimism/) repeats the pattern of a missing check. In all three cases, the code does what the formula says, and the formula describes a situation that no longer exists.

None of those losses requires a chain of exploits. It is enough for the system's state to fall outside the expected range and for no one to validate it before the money leaves the pool.

There is another uncomfortable reading: the market had gone months without volume and was still accepting deposits, so the flaw was not hidden in a forgotten corner of the code: it was in a product no one had closed.

## Where the money went and governance

Kava's update does not promise to return the funds. The team talks about accelerating the protocol's liquidation and putting the definitive shutdown to a vote, which leaves users with open positions in a situation that depends on the treasury and the community. In DeFi, that response is decided in a forum, not in a court.

For anyone who had deposits in HARD Lend, the practical priority is to check whether their market was among those affected and whether they can withdraw. Healthy markets usually remain operational during an orderly shutdown, and those on the attack path do not. The distinction depends on each pool, and it is worth checking before assuming everything is frozen.

The precedent of other shutdowns is not encouraging. When a protocol decides to wind down, the withdrawal schedule stretches out, interest rates freeze and users who arrive late compete for the liquidity that remains. Governance can approve compensation, and it can also fail to approve it without anything in the contract requiring otherwise.

The other unknown is the token itself. If the shutdown removes KAVA's main use within its ecosystem, demand depends on what remains to be built around it. The network will keep running, but an asset without a product is an asset with fewer reasons to be bought.

Users who decide to wait for the vote should read the full proposal before voting. A poorly executed shutdown leaves the last depositors with the leftover liquidity, and that is usually the worst part of the split.

## HARD Lend delivers another "hard" lesson on credit in DeFi

Decentralized credit replicates part of the banking business with one important difference: there is no regulator reviewing the books every quarter. The accounting runs in public, and that should make errors more visible, but visibility is not the same as oversight, because no one watches all markets at once when only one has volume.

The other lesson is about why zombie markets exist. A pool with inactive BUSD still generates interest for whoever uses it, and closing it requires a vote that no one proposes because there is no user complaint. The cost of keeping it open is not visible until someone exploits it, and that calculation is what teams will have to change.

Kava's decision to shut down the protocol instead of repairing it is a conservative response and, in this case, a defensible one. A market that lost 53.9 million KAVA to a formula cannot be fixed with a patch if no one trusts the rest of the system.
