My generation was blocked or flagged
Both our safety checks and the underlying model providers can block content. Rephrasing usually resolves a false positive.
Content can be blocked at two points: by our safety checks, and by the lab whose model you are using. Each provider applies its own filters, which is why the same prompt can be refused on one model and accepted on another.
Common reasons
- Sexual content involving real people without consent, or anything involving minors — this is a hard block with zero tolerance
- Named public figures, or recognisable copyrighted characters and brands
- Graphic violence or self-harm
- Wording that reads as an attempt to impersonate or deceive
If you think it was a false positive
Filters work on wording as much as intent, so a legitimate prompt can trip one. Rephrase the specific phrase that is likely to have caused it, describe a subject generically rather than by name, and remove references to brands or characters. Trying a model from a different provider often works too, since the filters differ.
Some things will not work regardless of phrasing, and should not. Our terms prohibit non-consensual deepfakes, sexual content of real people without written consent, impersonation, harassment and any CSAM. Accounts can be suspended for attempting these.
Likeness and voice
If you are working with a real person’s face or voice, you must have documented written consent from them. See our policy on likeness and voice for what is required.
Last reviewed against the live product in August 2026.