AI Safety & Responsible AI
Safety filter
A control that blocks, alters, or flags potentially unsafe AI inputs or outputs.
Example
A filter blocks a request for malware instructions before it reaches the model.
Why it matters
It helps stop certain unwanted material before it reaches a user or connected action.
You'll hear it
“The filter blocked a harmless quotation in this document.”
Can you control it?
Sometimes
Sometimes. Some products or workflows let you influence a safety filter, while others keep it fixed or manage it automatically. Check the settings and documentation for the tool you are using.
Common questions
- Does a filter catch every harmful request?
- No. New wording or unfamiliar situations can evade its checks.
- Can the filter act after an answer is written?
- Yes. Some systems check requests, replies or both.
- Does a filter's approval prove the facts are correct?
- No. Safety screening and factual verification are different tasks.