A Note on Language: Three Terms I Use When Talking About AI Governance.

A lot of AI governance conversations drift because we’re using the same words to mean different things.
Before I go deeper on drift, monitoring, or “model decline,” here are three terms I use consistently.
Behavioral Drift
Changes in how a system is used, relied on, or interpreted over time, even when the underlying model appears unchanged.
Continuous Assurance
Producing durable evidence as behavior occurs, rather than reconstructing it after an incident.
Reliance Formation
The point at which systems begin to be depended on operationally, often before governance and evidence mechanisms are in place.
I’ll use these terms consistently going forward so discussions stay grounded in time, behavior, and evidence.