Researchers from Elliptic, IBM Watson, and MIT have utilized artificial intelligence (AI) to detect money laundering on the Bitcoin blockchain. In 2019, Elliptic, in collaboration with the MIT-IBM Watson AI Lab, conducted research demonstrating how a machine learning model could spot Bitcoin transactions made by illicit actors like ransomware groups or darknet marketplaces. The partners have now released new research using advanced techniques on a larger dataset of nearly 200 million transactions. Instead of pinpointing transactions by illicit actors, the machine learning model was trained to identify “subgraphs” – chains of transactions indicating bitcoin being laundered. By focusing on these subgraphs, the researchers were able to analyze the “multi-hop” laundering process more broadly. Testing their technique with a crypto exchange, the researchers found that out of 52 predicted money laundering subgraphs ending with deposits to the exchange, 14 were linked to users flagged for money laundering. The team noted that less than one in 10,000 accounts were flagged on average, indicating strong performance of the model. Elliptic stated that this research showcases the effectiveness of using AI methods on blockchain data to uncover illicit wallets and money laundering patterns that were previously hidden. The transparency of blockchains allows for effective AI-based financial crime detection, showing that cryptoassets are more amenable to such detection compared to traditional financial assets.