China researchers develop AI crypto laundering detector with nearly 90% accuracy

Researchers at the People’s Public Security University of China have developed an AI framework that identified illicit Bitcoin transactions with 89.4% overall accuracy on a widely used public benchmark, adding a more specific technical and performance picture to earlier reports on the project. The study, published in May in the Journal of Intelligence and highlighted by the South China Morning Post, tested the model on the Elliptic Bitcoin dataset, which contains 203,769 transaction nodes and 234,355 edges. The system combines dynamic graph neural networks, a memory module and large language models. The researchers said that design helps the model track changing relationships between transactions, reference historical patterns of illicit activity and produce both risk scores and natural-language explanations for its classifications. It achieved 89.1% precision and 64.5% recall for illicit transactions, meaning flagged transactions were usually suspicious but some illicit flows still went undetected. The work lands amid a broader Chinese crackdown on crypto-linked financial crime. The South China Morning Post reported that Chinese researchers at the National Police University under the Ministry of Public Security developed a detection algorithm framework to identify illegal money laundering and economic crimes involving virtual assets such as Bitcoin with around 90% accuracy. China’s Supreme People’s Procuratorate said 3,259 people were prosecuted in 2025 alone on suspicion of money laundering involving virtual assets and underground finance. The study focuses specifically on Bitcoin, underscoring its longstanding role in illicit transaction monitoring.