Loading market data...

AI Trading Firm Triggers $60M in Liquidations After 19% Mark Price Drop in Korea

AI Trading Firm Triggers $60M in Liquidations After 19% Mark Price Drop in Korea

A single pre-market trade in South Korea sent a crypto asset's mark price tumbling 19% this week, triggering roughly $60 million in liquidations across multiple exchanges. The company behind the AI-driven trading strategy confirmed it will cover all losses from the incident, but insisted its oracle system worked exactly as designed.

The trade and the drop

The event unfolded during pre-market hours on a Korean exchange. A large order from an automated trading firm caused the mark price—a reference used by derivatives platforms to calculate liquidations—to fall sharply. The 19% move was enough to cascade through leveraged positions on several venues, wiping out longs before the market could stabilize.

Company response

The firm responsible for the trade issued a statement saying it would make all affected users whole. It did not disclose the exact number of accounts hit or the total payout amount beyond the $60 million figure. The company also defended its oracle, which feeds price data to exchanges, calling its performance “exactly as designed.” That phrasing has drawn scrutiny from traders who argue a system that allows a single trade to trigger a 19% swing is itself a risk factor.

Oracle debate

Oracles are supposed to provide reliable price feeds even under unusual market conditions. The company's insistence that nothing went wrong on its end raises questions about how exchanges set their mark price bands and liquidation thresholds. If a single order can move the reference price by nearly a fifth, the safety margins may be too thin—or the liquidity too shallow. Either way, the $60 million in losses is now the firm's problem to cover, not the users'.

What comes next

The affected exchanges are expected to review their pre-market trading rules and mark price calculation methods. The AI trading firm has not said whether it will adjust its own algorithms or limit order sizes during low-liquidity windows. For now, the promise to cover losses has contained the immediate fallout, but the incident has revived a familiar debate: how much automation should be allowed when the market is thin.