The researchers have also published the training data on the Google-owned machine learning and data science community site Kaggle. This move is expected to inspire more AI-focused research into bitcoin money laundering. However, the use of AI-based money-laundering investigation tools could raise new ethical and legal questions, particularly if they are used as criminal evidence. Despite this, the researchers believe their work will enable a new way to detect patterns that reveal financial crime.
Key takeaways:
- Researchers from Elliptic, MIT, and IBM have developed an AI model that can detect money laundering on Bitcoin's blockchain by identifying patterns of transactions between known criminal entities and cryptocurrency exchanges.
- The team has released an experimental version of the AI model and a 200-million transaction dataset, which they describe as the largest of its kind ever made public.
- The AI model has proven to be effective, correctly identifying 14 out of 52 suspicious accounts in a test with a cryptocurrency exchange.
- Despite the success, some experts warn that AI-based money-laundering investigation tools could raise ethical and legal questions, especially if used as criminal evidence, due to their "black box" nature.