What is model drift and its implications for prompt maintenance?

• A. Data Drift Is The Only Issue; Prompts Don’t Need Maintenance
• B. Hardware Failures Are The Main Concern For Prompts
• C. Model Updates Change Behavior, Prompt Maintenance Required
• D. Models Never Drift After Deployment

Answer: (C)

Model drift means the model’s behavior can change over time, especially after updates, retraining, or new data being incorporated. This matters for prompts because the same prompt that guided the model to produce a desired result can start behaving differently once a new model version is deployed or the training data shifts. To keep outputs aligned with goals, prompts need maintenance: you adjust wording, add system messages or constraints, and sometimes revise the prompt template so the intended intent still leads to the correct behavior with the updated model. This also often involves versioning prompts, running tests, and monitoring outputs to catch regressions early. Hardware issues aren’t the main factor here, and while data drift matters, the focus in this context is that changes to the model itself drive changes in how prompts behave, so ongoing prompt maintenance is necessary.