Ajna Protocol, a lending platform designed to operate without price oracles, reportedly lost approximately $775,000 in ETH after attackers abused the platform's internal liquidation accounting rather than compromising third-party price feeds.

The exploit affected multiple liquidity pools, including syrupUSDC, wstETH, rETH, cbETH, WBTC, WETH/USDC, and sDAI. The incident raises questions about a core element of Ajna's design philosophy: the absence of oracles and governance in favor of a self-pricing market. The attacker appears to have profited from that very assumption. The case is being watched closely because oracle removal has been marketed across DeFi as a security upgrade — an argument Ajna itself has made — and a breach of the internal logic that replaces the oracle tests the premise directly.

The oracle Ajna deliberately left out

Most lending protocols rely on an external service such as Chainlink to determine collateral prices. Ajna intentionally does not. Its white paper describes the protocol as "a non-custodial, peer-to-peer, permissionless lending, borrowing and trading system that requires no governance or external price feeds to function."

Instead, lenders set the rates at which they will lend by depositing funds into fixed-size "buckets," and the protocol's contracts determine when a loan is liquidated. To initiate a liquidation, the party starting the process must post a liquidation bond, which imposes a financial penalty if a liquidation is carried out without justification. The trade-off is that the complexity an oracle would have handled externally moves inside the protocol's own contracts — which is exactly where this exploit reportedly landed.

Security firm MixBytes explained the rationale for removing the oracle: "a significant portion of attacks on DeFi protocols stem from oracle prices manipulations, errors in configuration and access control issues."

Ajna's remedy was to eliminate that attack surface and trust the pool's own operations. A warning from Defimon suggests the attacker focused precisely on this internal mechanism — extracting funds by manipulating liquidation accounting rather than compromising an external price feed.

An hour of warning that went unanswered

Defimon stated that it detected a prepared attack more than one hour before the first exploit transaction and notified Ajna via the project's Discord chat (X post). The protocol had not yet been secured when the assault began. The gap between warning and exploitation underscores a recurring operational weakness in DeFi: even credible alerts depend on a team's ability to pause or patch contracts in time, and permissionless protocols often have limited centralized kill-switches by design.

The attacker then moved across numerous pools. According to the report, the syrupUSDC pool accounted for approximately $173,700 of the roughly $775,000 total loss.

The loss is significant relative to Ajna's size. At the time, DefiLlama data showed total value locked (TVL) for Ajna V2 at about $206,000, with active loans of around $418,000 and a 30-day TVL change of -54.2%. The reported loss from the attack therefore exceeded the protocol's TVL. DefiLlama's live figures have changed since then.

Break the code, or make it believe something impossible

The larger question is whether the attacker broke Ajna's code or tricked the system into treating false data as valid. The available evidence points to the latter.

Ajna's published audit history includes past findings related to "take" computations during liquidation and instances where bucket-state accounting was handled incorrectly, among other issues. Those problems were deemed fixed, but they demonstrate a history of complexity in the liquidation and accounting logic.

The pattern is a familiar one. Cryptopolitan previously reported on Moonwell, where an attacker used roughly $7 million to pump the illiquid MAMO token eightfold, then borrowed nearly $10 million of real assets and ultimately escaped with close to $6 million.

According to Nethermind, such attacks work as follows: "they force the contract to calculate a distorted price and exploit it before the transaction ends." Ajna removed the oracle, but its contracts still have to trust their own calculations.

V2's architecture becomes the real story

Ajna V2 currently shows about $450,000 in TVL against $30,200 in active loans, with TVL down 17.1% over 30 days.

A notable analytical statistic is that active loans amount to only about 6.7% of reported TVL. That sharpens the central investigative question: is the suspected exploit affecting outstanding debt accounting, deposited liquidity, or both? In other words, "What assumption did V2 introduce that an attacker could turn into money?" A post-incident audit or detailed technical disclosure identifying the exact accounting flaw will determine whether the vulnerability is specific to V2's bucket and liquidation logic or points to something deeper in the oracle-free model itself.

A small pool in a record year for exploits

The $775,000 loss is modest compared with the largest crypto hacks of 2026, but it fits a broader pattern.

TRM Labs counted 207 hacks in the first half of the year — the highest number it has recorded in a six-month period — with the typical incident costing about $219,000. More than 100 of these involved smaller smart-contract exploits. Infrastructure and operational compromises represented only about 15% of incidents but accounted for roughly 76% of total losses.

Ajna illustrates a different dimension of the security problem: losses do not have to stem from spectacular exchange breaches or compromised private keys. They can emerge from the assumptions buried inside DeFi's increasingly complex lending logic. For smaller protocols, where a six-figure loss can exceed total deposits, that lesson carries proportionally greater weight.