South Korea’s Financial Supervisory Service (FSS) has introduced a real-time AI surveillance platform designed to identify potential cryptocurrency market abuse by analyzing trading activity, news, exchange announcements and online discussions.
The system combines generative AI and machine learning to detect unusual market behavior, including rapid price manipulation, wash trading, coordinated trading and misleading promotional campaigns. AI-generated alerts will still be reviewed by human investigators before the regulator decides whether to conduct a deeper analysis or launch a formal investigation.
AI monitors abnormal price and trading activity
The FSS said the new platform automatically scans crypto markets for unusual price and volume movements, then compares those patterns with characteristics observed in previous market-manipulation cases.
The system can identify short-lived price surges and unusual market conditions, including situations where an asset rises sharply while deposits or withdrawals are restricted. It builds on an AI-based surveillance tool introduced earlier this year to help investigators identify suspicious orders and trading periods.
For potential wash trading and coordinated activity, the FSS also uses machine-learning models alongside Benford’s Law, a statistical method that examines the distribution of leading digits in numerical datasets. Unusual patterns are used to identify assets and time periods that may warrant closer examination.
AI also scans news, social media and chat rooms
The surveillance system does more than monitor exchange data. When an asset experiences an unusual move, generative AI searches news reports and exchange announcements for legitimate explanations, such as listings, network upgrades or other significant developments.
If no clear catalyst is found, investigators can request additional trading and account information from the relevant exchange.
The FSS is also expanding surveillance to online content. The system converts text, video subtitles and audio from platforms such as YouTube, online forums and messaging channels into searchable data. It then looks for potential false claims, coordinated promotion and trading activity that could indicate manipulation.
The regulator is particularly interested in cases where individuals allegedly trade ahead of their followers, distribute misleading information or coordinate promotional campaigns designed to attract retail investors.
Despite the expanded automation, human investigators remain responsible for determining whether an alert justifies further action.
South Korea has investigated more than 40 suspected cases
The initiative follows the implementation of South Korea’s Virtual Asset User Protection Act in July 2024. The legislation introduced stronger safeguards for crypto users and gave authorities additional tools to address market abuse, including insider trading, wash trading and price manipulation.
South Korean authorities have examined more than 40 suspected unfair-trading cases since the law came into effect. More than 30 cases have reportedly been referred to investigative agencies, involving multiple suspects and significant alleged illicit gains.
Crypto exchanges have also introduced additional controls. Members of the Digital Asset Exchange Alliance, including Upbit, Bithumb, Coinone, Korbit and Gopax, have strengthened monitoring of API keys to address risks associated with unauthorized access, coordinated trading and potential manipulation.
The country is also preparing broader digital-asset legislation covering areas such as stablecoins, exchanges, disclosures, internal controls and operational resilience.
AI surveillance raises accuracy and accountability questions
South Korea’s approach reflects a wider global effort to use AI for financial-market supervision.
In the United States, regulators have already used AI to identify suspicious trading patterns and potential regulatory violations. However, government reviews have emphasized the importance of human oversight, with specialists expected to assess automated alerts before regulators take further action.
The FSS is moving a step further by incorporating generative AI directly into market-surveillance workflows. That could significantly reduce the amount of data investigators need to process manually, but it also introduces concerns around false positives, unreliable outputs and the possibility of automated systems misinterpreting legitimate market activity.
The distinction is particularly important when an AI-generated alert could eventually contribute to regulatory or legal action.
Cross-exchange and on-chain monitoring next
The FSS plans to expand the platform beyond individual exchange datasets. Future upgrades are expected to include cross-exchange fund-flow analysis and on-chain transaction monitoring, potentially giving investigators a broader view of how suspicious activity moves across centralized platforms and blockchain networks.
For South Korea, the objective is clear: use AI to identify potential market abuse faster while keeping final investigative decisions in human hands.
The next challenge will be ensuring that increasingly sophisticated surveillance technology improves enforcement without turning automated signals into conclusions before the underlying evidence has been independently verified.
