content harm filters in Microsoft 365 Copilot: what does this AI read, and can you turn it off?
Checked on October 22, 2026 by verification chain vchain-1
This AI is on by default; it can be turned off.
The facts, edition by edition
| All editions | |
|---|---|
| What the AI does | Sort “content harm filters to identify harmful content in prompts or generated responses (such as Hate & Fairness, Sexual, Violence, and Self-harm)” [1] dated copy |
| Model | Not yet verified |
| What it reads | Other data “content harm filters to identify harmful content in prompts or generated responses (such as Hate & Fairness, Sexual, Violence, and Self-harm)” “content harm filters to identify harmful content in prompts or generated responses” [1] dated copy |
| Where the data goes | Not yet verified |
| Trains on your data | Not yet verified |
| On by default | Yes “The default moderation level is High.” [2] dated copy |
| Can be turned off | Yes “Microsoft offers certain content filtering controls for admins and users where filtering of potentially harmful or sensitive content in Microsoft Copilot Chat doesn't meet their needs.” [3] dated copy |
| Billing | Not yet verified |
| Feature type | Not yet verified |
| Frameworks the catalog links to it | Not yet verified |
All editions
- What the AI does
- Sort “content harm filters to identify harmful content in prompts or generated responses (such as Hate & Fairness, Sexual, Violence, and Self-harm)” [1] dated copy
- Model
- Not yet verified
- What it reads
- Other data “content harm filters to identify harmful content in prompts or generated responses (such as Hate & Fairness, Sexual, Violence, and Self-harm)” “content harm filters to identify harmful content in prompts or generated responses” [1] dated copy
- Where the data goes
- Not yet verified
- Trains on your data
- Not yet verified
- On by default
- Yes “The default moderation level is High.” [2] dated copy
- Can be turned off
- Yes “Microsoft offers certain content filtering controls for admins and users where filtering of potentially harmful or sensitive content in Microsoft Copilot Chat doesn't meet their needs.” [3] dated copy
- Billing
- Not yet verified
- Feature type
- Not yet verified
- Frameworks the catalog links to it
- Not yet verified
To turn it off
- All editions: Yes “Microsoft offers certain content filtering controls for admins and users where filtering of potentially harmful or sensitive content in Microsoft Copilot Chat doesn't meet their needs.” [3] dated copy
Frameworks the catalog links to it
Linked frameworks will appear once the reference set is published.
- All editions: Not yet verified
Sources
- Data, Privacy, and Security for Microsoft Copilot, Microsoft, October 22, 2026
- Knowledge sources summary, Microsoft, October 22, 2026
- Application card: Microsoft Copilot (for organizations), Microsoft, October 22, 2026
Change log
No change since first publication.
This fact sheet says nothing about your setup: the free readout, in French for now, reads it in your tools.