A company whose AI-driven trading system triggered approximately $60 million in cryptocurrency liquidations has committed to covering all user losses resulting from the incident. The event occurred when a single pre-market trade in Korea caused mark prices to plummet 19%, triggering widespread forced liquidations across affected platforms.
The company maintains that its oracle system functioned exactly as designed, according to reporting from CoinDesk. The statement suggests the liquidations stemmed from a pricing discrepancy between the oracle's data feed and actual market conditions, rather than a malfunction in the oracle mechanism itself.
Oracle systems serve as critical infrastructure in decentralized finance, providing external price data that smart contracts use to determine collateral values and trigger liquidations. When oracle prices diverge significantly from real market prices—whether due to thin liquidity on reference exchanges, manipulation, or latency—protocols can execute liquidations based on inaccurate valuations. The 19% mark price drop on a single trade indicates potential thin order book conditions during pre-market hours, allowing a single large order to temporarily distort the price feed.
The firm's decision to fully compensate affected users represents a significant assumption of liability in an industry where protocol teams often disclaim responsibility for oracle-related losses. This response distinguishes the incident from past oracle failure cases, where users typically bore losses without recourse. The move may reflect the company's confidence in its technical design and desire to preserve user trust, or conversely, recognition that the oracle's design parameters may have created predictable vulnerability to specific market conditions.
The incident highlights ongoing challenges in oracle architecture, particularly regarding pre-market and low-liquidity periods when price discovery mechanisms are most fragile. Exchanges and protocols continue grappling with trade-offs between oracle responsiveness—which enables efficient capital usage—and robustness against temporary price distortions. The 19% single-trade impact suggests the oracle may have weighted a single exchange or trading venue too heavily, lacked sufficient outlier detection, or operated with insufficient staleness checks for off-hours trading.
Industry observers will likely scrutinize whether the company's reimbursement commitment establishes precedent for similar incidents, and whether oracle design standards evolve to better account for pre-market liquidity dynamics. The case adds to a growing body of examples demonstrating how automated systems can amplify individual trades into systemic consequences when price feed architectures encounter edge-case market structures.