Automate NSFW moderation.

An upload pipeline cannot rely on someone opening every file. Is this NSFW? Your application needs a consistent answer for each incoming item, whether it is a profile photo, a community attachment or a newly generated image. Call the API before publication and use the returned scores in your own approval rules.

AI image products can check outputs before delivery. Communities can screen attachments before they enter a feed. Marketplaces can review new product images, and dating platforms can apply the same photo policy across profile submissions.

Start with one integration and reuse it across image, video and text inputs. Keep the moderation decision associated with your own content ID so your application can approve, hold or reject the correct item.

Create an account, generate an API key and use the playground to test representative inputs. Send the key from your server, never from a public client. Images accept a public HTTPS URL or base64; video accepts a direct public HTTPS URL; text accepts a string.

Call the API whenever your platform receives a new upload or produces an output. Give each item its own idempotency key and reuse that key when retrying the same content. Pace requests to the account rate limit and honor Retry-After on a 429 response. A completed request with the same key returns its existing result without another charge.

Read the verdict and four category scores, then apply your product policy. Publish accepted items, hold uncertain cases for review and handle rejected content in your own application. Store the result alongside the item so moderation does not depend on a person revisiting the playground.

The API returns Sexual, Nudity, Violence and Gore scores. Those categories let your application distinguish different risks instead of treating every flag as the same problem. A score is a risk estimate, not a complete understanding of the content or its context.

Suggestive content can imply sexuality through clothing, pose or framing without depicting explicit activity. Nudity is a separate consideration: a medical image, a classical painting and an explicit photograph may require different treatment under your policy. Asking “is this image NSFW?” is only useful when the answer is tied to the audience and the rules of your product.

Use representative accepted and rejected examples to choose a threshold. Lower thresholds flag more items and create more review work; higher thresholds can miss more problematic content. Evaluate false positives and missed cases together. The same NSFW detector can support different policies, but a threshold does not replace a written policy.

Check the generated output, not just the prompt. An AI image service can produce an unexpected detail even from an ordinary request. Place the check after generation and before an image enters a gallery, customer download or public feed.

An AI NSFW detector can misread stylized anatomy, unusual proportions and abstract scenes. Include photographs, anime, drawings and the styles your product actually produces in your evaluation set. Test the final resolution and crop used in production.

Separate uncertain results from clear policy violations in your own workflow. A review queue allows your team to inspect ambiguous items without blocking every borderline illustration or automatically publishing every low-scoring output.

Submitted content is sent to the service for detection. Temporary file copies are deleted when processing completes or fails. We do not maintain a submitted-file library, publish customer files, manually view them as part of detection or use submissions to train our own models.

Detection results and operational records are separate from the submitted content. We retain scores, request identifiers, processing status and credit records to support usage history, retries and billing. The original file and text are not stored in those records.

Service providers process content under their applicable terms. Our deletion statement covers copies controlled by this application, not a blanket retention promise for every provider. Review the Privacy Policy before connecting a production workload, and submit only content your organization is entitled to process.

NSFW means “not safe for work.” In a moderation policy, it usually signals sexual content, nudity or graphic material that should not appear unexpectedly. The API also reports Violence and Gore as separate risks.

Your platform determines what happens after a flag. An art marketplace, a general-audience community and a workplace product can have different acceptance rules. Use the NSFW scanner as a classification input, then make your own allow, review or reject decision.