South Korea Deploys Real-Time AI System to Detect Crypto Manipulation

Real-Time AI Surveillance System Launched

South Korea’s Financial Supervisory Service (FSS) has deployed a real-time artificial intelligence system designed to scan trading data, news, and online content to identify suspected cryptocurrency price manipulation. According to an FSS announcement, the platform integrates generative AI and machine learning to automate the detection of abnormal price and volume spikes, wash trading, collusive activity, and false promotional claims.

The new setup builds upon an algorithm introduced in January and connects initial alerts, supporting data, and preliminary reviews into a single workflow. The system flags assets showing unusual volatility and compares them against patterns from past investigations, including the "racehorse" pattern of sharp short-term moves and the "cage" pattern involving steep price rises while deposits or withdrawals are restricted. For potential wash trading or coordinated trading, the FSS utilizes machine-learning models alongside Benford’s Law.

Screening News and Online Content

Once a token experiences an unusual market move, generative AI reviews relevant news and exchange announcements to determine if a verified catalyst, such as a listing notice or network update, explains the activity. If no clear reason is found, the regulator may request detailed order and account data from the relevant exchange.

The surveillance mechanism also evaluates digital promotion by converting text, video subtitles, and audio from YouTube, online forums, and messaging chat rooms into text. The platform scans this material for suspected front-running, false information, or coordinated buy recommendations designed to induce unfair retail trades.

Human Oversight and Enforcement Context

Despite the automation, human investigators remain strictly responsible for reviewing AI-generated reports before opening detailed analyses or formal investigations. Future updates to the surveillance platform are slated to include cross-exchange fund-flow analysis and on-chain transaction tracking.

The deployment follows enforcement efforts under South Korea’s Virtual Asset User Protection Act, which went into effect on July 19, 2024. Over the law's first two years, Korean authorities examined more than 40 unfair-trading cases, referring over 30 cases to investigative agencies and identifying 25 suspects with average calculated unlawful gains of roughly 1.4 billion won ($940,000) per case.