Coinbase: Shortening Fraud Response Time with New AI-Driven Rule Engine

CryptoNews reports that Coinbase has rebuilt its anti-fraud system by tightly integrating machine learning models with high-speed rule engines, reducing its response time to new scam patterns from days to a few hours. With TRM Labs warning that crypto fraud has become an AI-driven industry worth several tens of billions of dollars annually, Coinbase’s upgrade is intended to address the surge in AI-driven scams. The company uses a dual-track strategy: the models handle long-term defense, while the rules provide rapid response. All functions are integrated into a unified framework to detect new types of fraud and feed them back to the models to strengthen overall protection. Coinbase says it has transformed what used to be a manual and slow rule-creation process into a data-driven automatic recommendation system, boosting rule backtesting performance by more than 10 times and allowing it to test and roll out new safeguards more quickly. According to Coinbase, the system now uses machine learning to recommend rule parameters, aiming to “reduce false positive rates while combating fraud and minimizing impact on legitimate users.”