A company at the center of an AI-driven trading incident that triggered roughly $60 million in cryptocurrency liquidations has pledged to cover all losses suffered by affected users, even as it maintains that its underlying systems performed as intended.
What Happened
The disruption stemmed from a single pre-market trade executed in Korea, which sent the relevant mark price plunging 19% in one move. That sharp, sudden drop rippled through connected trading systems, cascading into liquidations that ultimately wiped out around $60 million in positions.
The mark price is a critical reference figure used to value open positions and determine when leveraged traders get liquidated. When it swings violently on a thin or isolated trade, the effects can quickly spread to accounts that had no direct involvement in the transaction that caused the movement.
A single pre-market trade was enough to crater the mark price by 19% and set off a chain reaction of forced liquidations.
The Company's Response
Despite the fallout, the company insists its price oracle — the mechanism responsible for feeding market data into the platform — functioned exactly as it was designed to. In other words, the firm frames the episode as an edge-case market event rather than a technical failure of its infrastructure.
Even so, the company has committed to making affected users whole, promising to cover the full extent of the losses tied to the liquidations. The move appears aimed at preserving user trust following an incident that highlighted how fragile pricing can be during illiquid trading windows.
Key points from the incident include:
- A 19% collapse in the mark price triggered by one pre-market trade
- Roughly $60 million in crypto positions liquidated as a result
- The company's stance that its oracle behaved correctly
- A pledge to reimburse all impacted traders
The episode underscores ongoing concerns about how automated trading systems and oracles respond to outlier events, particularly in markets with limited liquidity where a lone transaction can distort pricing across an entire platform.