The country’s financial regulator is upgrading its surveillance platform to spot suspicious trading patterns in under five minutes — a window that human investigators could never reliably match. The Bithumb incident made the case impossible to ignore.
Image created by ChatGPT
South Korea’s Financial Supervisory Service (FSS) has announced its 2026 work plan, with AI-driven crypto surveillance at its centre. The regulator is upgrading VISTA (Virtual Assets Intelligence System for Trading Analysis) to automate the initial detection of suspicious trading activity, replacing a process that previously relied heavily on human investigators scanning data after the fact.
The upgraded VISTA system uses a sliding-window grid search method — dividing trading data into overlapping time segments of varying durations and scanning each for anomalies, without investigators needing to specify where misconduct might have occurred first. In internal testing, the upgraded AI detected all known manipulation periods from past completed investigations, and additionally flagged intervals that human analysts had previously missed.
The timing matters. In recent months, a high-profile operational failure at Bithumb, South Korea’s second-largest crypto exchange, exposed just how quickly things can go wrong in digital asset markets and how difficult it is to respond in time. The FSS has since confirmed it will investigate potential price manipulation linked to the incident, making the 2026 oversight plan both a strategic upgrade and a direct policy response.
The FSS has been specific about which trading behaviours it considers highest-risk. The plan distinguishes several named tactics that the AI system will be trained to detect:
Common tactics used in crypto market manipulation. Image created by ChatGPT
A key feature of the upgraded VISTA is its ability to analyse text signals alongside trading data — scanning social media and communications for unusual promotional spikes that correlate with price movements, in an attempt to catch narrative-led manipulation before it distorts the market. These capabilities can be useful in situations similar to the alleged coordinated crypto attack that recently affected Trump-backed USD1 stablecoin.
Perhaps the most significant policy development in the plan is not what the AI detects, but what regulators are considering doing with that information. The Financial Services Commission is exploring a mechanism that would temporarily suspend transactions or payments linked to suspected manipulation, with the aim of freezing illicit gains before they can be withdrawn or laundered.
This represents a shift from reactive enforcement, which is investigating after the fact, to proactive intervention. It is not yet finalised, and the FSS has acknowledged the significant governance questions it raises: what thresholds trigger a suspension, who authorises it, how quickly can false positives be reversed, and what protections exist for innocent parties caught in the net.
The FSS is not limiting its AI ambitions to digital assets either. The Korea Exchange is separately implementing an AI monitoring system for conventional stock markets, aimed at detecting share price manipulation earlier. The same AI infrastructure will also support early warning systems for financial fraud and voice phishing, by cross-referencing telecom data with financial transaction records. The stated goal is a unified single financial intelligence system approach across asset classes.
The 2026 surveillance push sits alongside the second phase of South Korea’s legislative framework for virtual assets. The FSS has formed a dedicated task force to develop the Digital Asset Basic Act, which will introduce disclosure requirements for token issuance, licensing standards for digital asset operators, and clearer approval procedures for stablecoin issuers. The regulators are reportedly planning to allow local registered firms use up to 5% of their equity capital to buy digital assets and diversify their portfolios.
