Back to Glossary
ai-risksintermediateunit-7

Model Drift

Definition

When AI behavior changes over time, often unexpectedly.

In Plain English

Model drift is like an employee gradually changing how they do things without telling you.

Real-World Example

AI providers update models, and suddenly your prompts produce different results.

Why It Matters for Your Work

Drift can break applications that depend on consistent AI behavior.

Common Mistake

Not monitoring AI outputs over time. Drift happens gradually and can go unnoticed.

Related Terms

View AI
AI

Artificial Intelligence—software that can make predictions, generate content, or assist with decisions.

View Model
Model

The trained AI system that produces outputs based on inputs.

View QA
QA

Quality Assurance—testing to find and fix problems before launch.