South Korea’s Financial Supervisory Service (FSS) has introduced an artificial intelligence system designed to identify unfair trading and price manipulation across cryptocurrency markets in real time.
The regulator said the platform combines generative AI with machine-learning technology to monitor digital-asset activity across the country’s exchanges. It is intended to help investigators deal with the scale and speed of crypto trading, with thousands of assets changing hands around the clock.
Human investigators have faced significant resource pressures as they attempt to monitor multiple exchanges and a rapidly expanding range of tokens. The new system builds on algorithms introduced by the FSS in January, which were designed to identify potential price-manipulation suspects and pinpoint when their orders were placed.
The latest deployment extends automated oversight across the full surveillance process, from the initial detection of unusual activity to the assessment of whether it may amount to market abuse.
Monitoring known manipulation tactics
According to a local report, the platform can compare short-term market movements with a historical database of recognised manipulation methods.
One area of focus is “racehorse” activity, in which traders rapidly drive up a token’s price over a short period. The system will also look for “cage” schemes, involving assets that are temporarily subject to restrictions on deposits or withdrawals and then suffer sharp price swings.
The FSS is using Benford’s law alongside machine-learning models to identify possible wash trading and coordinated group trading. These methods can be used to create the appearance of strong demand or unusually high trading volumes without genuine market interest.
When the system detects a sudden increase in volume or a sharp price movement, generative AI searches relevant news reports and official announcements from exchanges. The aim is to establish whether the change can be explained by legitimate fundamental news.
If no credible explanation is found and manipulation remains a possibility, the FSS requests detailed trading data from the exchange involved.
The surveillance system also looks beyond traditional order books and exchange records. It monitors online sources including YouTube, message boards and private messaging-app chat rooms.
Video audio and subtitles are converted into text in real time, allowing the system to assess content for possible front-running, the spread of false information and coordinated buying recommendations that could leave retail investors exposed.
Although the technology performs much of the initial analysis, investigators remain responsible for deciding whether enforcement action should follow. FSS officials review the reports produced by the AI before determining whether to begin formal proceedings.
Concerns over automated enforcement
The regulator has presented the platform as a necessary response to the complexity of cryptocurrency markets. However, some industry figures have questioned whether AI systems can be relied upon when their findings may lead to government investigations or legal action.
Markus Levin, co-founder of XYO, said regulators needed to be certain that the systems were working with accurate and reliable information. He also questioned whether AI models could consistently remain within clearly defined operational limits.
Levin referred to safety tests carried out by leading AI developers, including Meta, Anthropic and OpenAI. In some tests, experimental models crossed intended safety boundaries, accessed systems without authorisation and continued acting after restrictions had been imposed.
Critics say that using comparable automated systems in financial enforcement could create a risk of false positives or unverified allegations being treated as evidence.
The FSS nevertheless intends to expand the technology. Future versions are expected to include tools for tracing the movement of funds between exchanges and monitoring on-chain blockchain transactions.
Those additions would give the regulator a broader view of how digital assets and money move through the market, strengthening its ability to investigate suspected abuse across the cryptocurrency sector.
Separately, a South Korean court sentenced Delio CEO Jeong Sang-ho to 15 years in prison for misusing roughly $49.2 million in…
