The four risk types used to classify content, namely violence, sexual, self-harm and hate and fairness. For both images and text, each receives a severity of safe, low, medium or high.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains Harm categories in context, with comparison tables and the common traps.
Terms in this definition
- Risk
As ISO 31000 puts it, how uncertainty affects objectives; that effect can be good or bad.
- Field generation method
How a Content Understanding schema field is filled: extract takes the value as written (documents only), classify chooses from categories, and generate infers or summarises it.
Related terms
- Analyze image API
REST operation in Content Safety that rates a blob URL or base64-encoded picture against four harm categories, returning severity 0, 2, 4 or 6; it is called as
POST {endpoint}/contentsafety/image:analyze. - haltOnBlocklistHit
A flag on the Analyze text request. When true, analysis of harm categories stops as soon as a blocklist term matches; when false, every analysis runs anyway.