ModelRefs / Refusal — AI Glossary
Refusal — AI Glossary
A model declining to respond to a request, citing safety, policy, capability, or ethical reasons. Also called rejection or model refusal.
Overview
Refusals are explicit safety behaviors trained into models to prevent harmful content generation. 'Over-refusal' (refusing benign requests) and 'under-refusal' (complying with harmful requests) are dual failure modes measured by safety benchmarks. Refusal calibration is a key alignment challenge—Claude's Constitutional AI, OpenAI's usage policies, and Anthropic's RSP all govern refusal thresholds.
Reference details
| Topic | safety |
|---|---|
| Also known as | rejection, model refusal, content refusal |
| Last reviewed | 2026-06-24 |
Related terms
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Refusal — AI Glossary.
Frequently asked questions
What is Refusal?
A model declining to respond to a request, citing safety, policy, capability, or ethical reasons.
Is Refusal the same as rejection?
Yes — rejection, model refusal, content refusal are common aliases for Refusal.
What concepts are related to Refusal?
Closely related concepts include guardrails, alignment, abstention.