A generation can be blocked in two places: by our own check before it starts, or by the model provider's safety filter while it runs. The message tells you which one it was.
Blocked before it started
The message starts with "This request was blocked" and names the rule: content involving anyone under 18, nude or sexual images of a real person from a photo, or extreme violence. No credits were taken.
The check reads whole words and some word combinations, so an innocent prompt can trip it. For example, a word about children next to a word about nudity or sex is always blocked, even when the intent is harmless, and nudity words are blocked whenever a photo is attached. Rewrite the prompt without that combination and try again.
Blocked by the model's safety filter
The job card says "Blocked by the safety filter". The provider refused the request, so there is no result. The first 3 of these in any 24 hours are refunded automatically; after that, credits for further rejections that day are not refunded. Recent charges and refunds are listed under Credit history on your account page.
Some models are stricter than others, especially with photos of real people. Trying a different model or a less ambiguous prompt often works.
Repeated blocks
After 5 blocked requests in 24 hours, counting both kinds, new generations are paused until the oldest of them is more than 24 hours old. The message tells you roughly how long that is. Accounts that keep trying to make prohibited content can be suspended from generating.
If you think it was a mistake
Contact us with the date, the tool or model, and what you were trying to make. Appealing a decision lists what to include. The rules themselves are in What you can create and the Terms.