A third of banks do not maintain logs for GenAI models

Only one firm responding to Risk Benchmarking’s inaugural Model Risk Management (MRM) study—a specialist European lender—says it registers and reviews a complete record of the production prompts and outputs for the generative artificial intelligence models it currently has in production.

Roughly half of banks, including three of the G-Sibs answering, say they keep logs, but use them principally for audit or post-event investigation, rather than continuous surveillance. A further five, including