South Korea’s markets regulator has rolled out a real-time AI surveillance system to spot suspected crypto price manipulation — combining generative AI, machine learning and live exchange feeds to automate what used to be a labor‑intensive detective job. What the FSS launched - On Aug. 20 the Financial Supervisory Service (FSS) unveiled a platform that continuously scans trading data, news, social posts and other online content to flag anomalous token moves. - The system builds on an algorithm the FSS began using in January, but now links the initial alert, supporting evidence and a preliminary human review into one end‑to‑end workflow. How it works - The platform monitors real‑time price and volume data and flags assets that exhibit abnormal behavior. Those events are compared against patterns derived from past investigations to prioritize likely cases of market abuse. Examples the FSS calls out include: - “Racehorse” moves — sharp token jumps over a short window. - “Cage” moves — rapid rises in price while deposits or withdrawals are suspended or restricted. - For suspected wash trading or coordinated activity, the system applies machine‑learning models alongside Benford’s Law (which identifies anomalous digit distributions in naturally occurring numeric data). - Generative AI sifts related news, exchange notices and other public signals to identify plausible catalysts (e.g., listings or network upgrades). If a spike lacks an explainable cause, the FSS can request detailed order and account data from the relevant exchange. - The platform also ingests complaints, tips and media reports, and converts audio/subtitles and forum/chat content from YouTube, forums and private messaging into text for analysis — searching for front‑running, false claims or orchestrated buy calls. Findings are compiled into a standard report for investigator review. Human oversight and limits - The FSS emphasizes that human investigators remain the decision‑makers: every AI report is reviewed by staff before any deeper analysis or formal investigation is opened. An FSS official said the tool is intended to help limited teams “respond quickly and efficiently” to increasingly complex unfair trading. Regulatory and market context - The rollout follows two years of enforcement under the Virtual Asset User Protection Act, which took effect on July 19, 2024. The law requires crypto firms to segregate customer assets and lets regulators act on insider trading, wash trading and price manipulation. - Over the law’s first two years, Korean authorities examined more than 40 suspected unfair‑trading cases, reported or referred 30+ cases to investigative agencies, identified 25 suspects and calculated average unlawful gains of about 1.4 billion won (~$940,000) per case. - Exchange operators have tightened controls, too. In May, members of the Digital Asset Exchange Alliance — Upbit, Bithumb, Coinone, Korbit and Gopax — introduced stricter API‑key rules: monitoring key sharing, implementing IP whitelists and invalidating keys after warnings. The FSS estimated API trading accounted for roughly 30% of domestic crypto turnover, raising concerns that shared keys can enable coordinated trading. Broader policy moves and comparison with the U.S. - On July 29 the Financial Services Commission outlined a consolidated bill that could fold 10 pending digital‑asset proposals into a single framework covering stablecoins, exchanges, disclosures, internal controls and system resilience. Meanwhile, the Virtual Asset User Protection Act remains the main law governing custody and market abuse as lawmakers negotiate a second‑stage framework. - Internationally, a May 2025 GAO review found U.S. federal financial regulators were using AI to identify risks and detect possible violations, but agencies generally did not treat model outputs as the sole basis for enforcement and — as of December 2024 — were not using generative AI for supervisory market‑oversight tasks. The FSS’s platform applies generative AI to supervisory workflows in ways U.S. agencies had not reported at that time. - A 2025 CFTC roundtable highlighted AI’s potential for real‑time detection of spoofing and wash trading but cautioned that fragmented exchange data complicates oversight. Under current U.S. law the CFTC can pursue fraud and manipulation in spot commodity transactions, but it does not yet routinely supervise spot crypto exchanges the same way it oversees registered derivatives markets; broader federal registration would likely require legislation such as the proposed CLARITY Act. Concerns and safeguards - Industry voices urge caution. XYO co‑founder Markus Levin warned regulators need reliable input data and clear operational limits to avoid false alerts or unverified allegations triggering needless inquiries. He pointed to reported safety lapses in experimental models from major AI labs as a reminder that automated findings must be checked. - The FSS has built human review into its process and says it will continue to rely on investigators to validate AI signals. What’s next - The regulator plans to add cross‑exchange fund‑tracing and on‑chain transaction tracking to the platform, though no deployment date was provided in the Aug. 20 announcement. Bottom line South Korea’s FSS has moved from pilot to a live, integrated AI surveillance workflow that blends machine learning, generative AI and human oversight to hunt market abuse across trading feeds and online chatter. The system tightens the regulator’s toolkit as enforcement and market‑level controls evolve — but experts stress that good data, clear limits and robust human checks are essential to avoid overreliance on automated signals. Read more AI-generated news on: undefined/news