Ajna v2 Exploit Drained $775,000 Without an Oracle Attack

Ajna v2 was exploited on August 29, 2026, in an attack that reportedly drained about $775,000 across seven Ethereum lending pools. The incident did not depend on corrupting an external price feed. Instead, available reporting points to manipulation of Ajna’s internal liquidation-accounting system—the same self-contained machinery the protocol uses to determine collateral, debt, and liquidation conditions. 1311

The public evidence is enough to identify the attack class and affected pools, but not to establish a definitive, transaction-level reconstruction of every state change. Claims about the attacker’s exact sequence should therefore be treated as provisional until a complete forensic analysis is published.

What happened in the Ajna v2 exploit?

The reported attack affected the syrupUSDC, wstETH, rETH, cbETH, WBTC, WETH/USDC, and sDAI pools. Aggregate losses were estimated at approximately $775,000, with one later breakdown putting the total at about $775,400. The syrupUSDC pool was reported as the largest individual loss at roughly $173,700. 1913

Ajna said it was investigating abnormal fund movements and advised users to:

• withdraw quote-token funds;
• repay outstanding loans; and
• stop interacting with the protocol. 111435

Those instructions reflected the immediate risk that further activity could interact with the accounting conditions being investigated.

The attack targeted liquidation accounting, not a price oracle

Ajna’s design intentionally avoids external price oracles. Rather than relying on a service such as Chainlink to establish collateral prices, its lending and liquidation process uses internal pool activity and bucket accounting. 333

In the normal liquidation process, debt and collateral are removed from buckets, which changes internal metrics including the Lowest Utilized Price, or LUP. Loans are evaluated using variables such as the borrower’s Threshold Price, the Neutral Price, and the pool’s LUP. 3437

That architecture removes one familiar DeFi attack surface, but it does not remove the need for reliable economic accounting. Reporting on the August 29 incident says the attacker manipulated this internal liquidation path so that collateral, debt, and liquidation-price treatment no longer represented the pools’ genuine backing. 1311

The important limitation is that the available reports describe the exploit at a high level. They do not independently verify the complete sequence of deposits, borrowing, liquidation actions, repayments, and withdrawals that produced the final loss. It is therefore more accurate to say that the attacker appears to have exploited internal liquidation accounting than to present an unverified transaction narrative as settled fact.

Why “oracle-free” did not mean attack-proof

The incident highlights a distinction that is easy to miss:

Removing external price feeds removes oracle risk; it does not remove pricing risk.

Ajna concentrated critical assumptions in its own market structure and liquidation rules. If those rules allow an attacker to move or exploit the protocol’s internal representation of collateral, debt, or liquidation prices, the resulting failure can resemble an oracle attack in economic effect even though no third-party feed was manipulated. 111

This is a broader lesson for DeFi lending systems. Security review must cover not only code-level vulnerabilities and external dependencies, but also whether the protocol’s accounting state continues to correspond to real economic value through unusual combinations of borrowing, liquidation, and repayment activity.

Immutability made emergency response harder

Ajna v2 was described as immutable, with no governance body, upgrade path, administrator key, or centralized mechanism capable of halting the affected contracts. 9

That design limits discretionary intervention and reduces certain forms of governance and upgrade risk. It also creates a sharp operational trade-off: if the deployed mechanism is flawed, the team may be unable to pause or patch it while an attack is in progress.

Defimon Alerts said it detected a prepared attack more than an hour before the first exploit transaction and warned Ajna through the project’s Discord channel. The monitoring firm also said the team failed to react. That is an account from Defimon, not an independently adjudicated finding, so the timing and response should remain qualified pending further evidence. 816

What the incident means for DeFi risk

The Ajna exploit was relatively small compared with the largest crypto thefts, but it illustrates why incident counts and protocol design matter as much as headline dollar losses. TRM Labs recorded 207 crypto hacks and exploits in the first half of 2026, the highest six-month total in its dataset; smart-contract exploits accounted for about 125 incidents. 1829

Separately, The Defiant reported approximately 70 DeFi exploits and $746 million in losses during the second quarter of 2026. Different trackers use different inclusion criteria, so these figures should not be treated as directly interchangeable. Together, however, they point to a busy threat environment for DeFi protocols. 23

Ajna’s case adds a specific warning: an architecture can avoid external oracles and still contain a critical internal economic dependency. Oracle-free design, immutability, and decentralization may reduce some risks while increasing others—especially when there is no emergency control available after abnormal accounting begins.