Trust & Safety API responses

OctoAI has multiple trust & safety capabilities to detect and potentially block NSFW content, including a prompt filter and image classifier.

Prompt filter

The prompt filter is applied to the image generation prompt. If a prompt is blocked, we’ll return a 400 code and the API response will include "message":"This request violates our Acceptable Use Policy."

Image classifier

The image classifier is applied to the generated image. If an image is classified as NSFW, the image won’t be returned and the API response will include removed_for_safety:true. Example response:

{
"images": [
{
"image_b64": "null",
"removed_for_safety": true,
"seed": 1
}
],
}

Images uploaded for fine-tuning are also assessed using the image classifier, and the API response will designate NSFW images within the rejected_assets field.