Drift Monitoring

Drift monitoring is the ongoing practice of checking whether an AI system's inputs or outputs have shifted away from the conditions it was built and tested under. When customer behavior, data sources, or the world changes, a model that once performed well can quietly become less accurate or less fair, often without any visible error. For compliance and risk teams, it is the evidence trail that shows a deployed system is still behaving as documented and approved.

1 item

ResearchGlobal2026-10-25

Only 18% of Banks Have Fully Embedded AI Governance, SAS Survey Finds

A survey of banking institutions by SAS and FT Longitude finds that only 18% have fully operational AI governance processes embedded across their organizations. Just 19% use real-time model drift monitoring, and 42% lack defined kill switches for AI models that exceed preset risk thresholds. The findings give compliance teams industry benchmarks to measure their own programs against as regulatory scrutiny intensifies.

banking model risk drift monitoring kill switches SAS FT Longitude model governance financial services