AI Safety & Responsible AI
Refusal
A model response that declines to complete a request because of safety or policy constraints.
Example
The assistant declines a dangerous request and offers a safer alternative.
Why it matters
It gives a system a way to avoid completing requests outside its permitted scope.
You'll hear it
“The assistant declined that request and suggested a safer route.”
Can you control it?
Sometimes
Sometimes. Some products or workflows let you influence a refusal, while others keep it fixed or manage it automatically. Check the settings and documentation for the tool you are using.
Common questions
- Does a refusal mean the AI is incapable of the task?
- Not necessarily. It may be restricted rather than unable.
- Can a refusal be mistaken?
- Yes. A harmless request can be interpreted too broadly or without enough context.
- Does a polite explanation make the refusal consistent?
- No. Similar requests can still receive different treatment across attempts or services.