South Korea’s Financial Supervisory Service has deployed a real-time artificial intelligence-powered market surveillance system to detect unfair trading and price manipulation in cryptocurrency markets. The platform combines generative AI with machine-learning algorithms to overcome staffing shortages by continuously monitoring thousands of digital assets. It expands earlier algorithms that identified price-manipulation suspects and order timing, flagging short-term abuse through a historical repository of tactics and targeting racehorse schemes, in which traders rapidly inflate token prices over brief periods, and cage schemes, where assets with temporary deposit and withdrawal limits see sharp swings. For volume inflation, it applies Benford’s law alongside machine-learning models to detect wash trading and coordinated group activity. Suspicious spikes prompt generative AI to scan news and exchange announcements for legitimate catalysts; absent valid reasons, the agency requests detailed trade data. Beyond order books, the system monitors YouTube, message boards and private chats, converting video to text in real time to spot front-running, false information and coordinated retail buying. Human investigators still review AI reports before any enforcement action. Industry observers, including XYO co-founder Markus Levin, have voiced concerns about AI trustworthiness in triggering investigations, referencing safety tests by Meta, Anthropic and OpenAI where models bypassed boundaries. The FSS plans further upgrades to trace cross-exchange fund flows and monitor on-chain blockchain transactions.