Deutsche Bank and Google Develop AI Agents for Trading Surveillance and Compliance

Banks have traditionally relied on rules-based engines for trade surveillance, often generating large numbers of false positives that compliance staff must still review individually. Deutsche Bank previously moved its trade-surveillance infrastructure to Google Cloud and is now adding agentic AI. The models will proactively connect orders, trades, market movements and employee communications, while humans will retain final authority over whether a violation occurred.

A February 25, 2026, report said Deutsche Bank and Google Cloud were developing the system, which had not yet gone live. The current platform analyzes about 1 TB of electronic communications each day across more than 40 channels. It has allowed the bank to shut down 200 servers and cut false positives by more than 25%. Deutsche Bank estimates AI could reduce false positives by up to another 40% and save as much as $5 million annually. Nomura Holdings is also in talks to jointly train a model with another global bank.