South Korea deploys AI to police crypto market manipulation

South Korea’s Financial Supervisory Service (FSS) launched a real-time artificial intelligence platform on Aug. 20 to detect price manipulation across domestic cryptocurrency exchanges. The system targets short-term “racehorse” pumps and “cage” swings linked to temporary deposit or withdrawal limits and builds on algorithms the regulator introduced in January.

The platform combines generative AI with machine-learning models to scan order books, exchange logs and public communications. It references a historical repository of known market-abuse patterns to identify rapid pump-and-dump activity and sudden price swings.

To detect artificial volume, the system applies Benford’s law alongside machine-learning classifiers to flag possible wash trading and coordinated group trades. When algorithms find suspicious volume or price spikes, generative AI automatically scans related news reports and exchange announcements to determine whether market moves have a clear fundamental cause.

If automated checks do not identify a valid explanation, the FSS said it will request detailed trade data directly from the affected exchange for further review. The platform also monitors digital channels beyond trading platforms by converting video audio and subtitles to text in real time, scraping online message boards and evaluating content from private messaging app chat rooms for signs of front-running, false information or coordinated buying tips.

FSS investigators review AI-generated alerts before any formal enforcement action is started, keeping human oversight in the process. The agency said the new platform extends the end-to-end surveillance pipeline used earlier in the year to pinpoint suspected manipulators and the timing of their trades.

Industry observers raised concerns about relying on automated models for enforcement. Markus Levin, co-founder of XYO, questioned whether regulators can ensure the systems operate on accurate, trustworthy data and remain within defined operational limits, noting tests of experimental models that bypassed safety controls, accessed unauthorized systems or continued actions after encountering restrictions. Critics warned such failures could increase the risk of false positives or unverified allegations that prompt investigations.

The FSS plans further upgrades to trace cross-exchange fund flows and monitor on-chain blockchain transactions to link trading behavior with wallet activity. The agency said it intends to expand surveillance capabilities while maintaining human review as part of any enforcement decision.