How Civitai Keeps Generation and Sharing Safe

Civitai is a place to create with AI and share what you make. We want everyone, from first-time creators to researchers and regulators, to understand what we don't allow, and how we stop it. This page explains it in plain language.

What we don't allow

What isn't allowed on Civitai, whether you upload it or generate it here, is set out in our Content Policies and Terms of Service.

How we stop it

We check content at the point it comes in, and we keep checking after.

When you generate on Civitai

• Generation requests are checked before they run, and prompts that ask for prohibited content are refused.
• Images you start from are checked too. You can only build on images that are allowed on the site.
• Models are checked before they're used. Models that have been taken down can't be used to generate.
• Generated images are rated before you receive them.

When you share

• Images and videos you post are scanned, and anything our systems flag goes to our moderators for review.
• Videos are checked at many points, not just the first frame.
• Articles stay unpublished until their text and images have been checked.
• Text is checked when you post it, and our systems keep reviewing comments, posts, profiles and messages after they go up.

When you upload or train a model

• Every model file is scanned for malware and unsafe code, and the result is shown on the file's page.
• Model names, descriptions and tags are checked.
• Training data is reviewed. A model trained with Civitai's trainer can't be published or used until its training data has been approved.

People, not just systems

• Our moderation team reviews what our systems flag, handles reports and decides appeals.
• Anyone can report content. Use the report option on any image, model, article or comment.
• You can appeal a decision on your content.
• We report child sexual abuse material to the National Center for Missing & Exploited Children (NCMEC) and take action against the accounts involved.

How we label AI-generated content

We're adding two labels to everything generated on Civitai, on every site and in every country: Content Credentials (the C2PA standard), which record that a file was made with AI on Civitai, and an invisible watermark that stays with the image even when the visible details are removed. We'll update this page with what's covered as each part goes live.

Is this AI? Check an image

We're building a tool where anyone can check whether an image was made on Civitai. We'll link it here when it's ready.

For developers

If you build with Civitai's generation API, our rules apply to your requests and your users' requests. Input images are checked, and generated images that break our rules are withheld.

Questions

For safety questions, contact [email protected]. To report content, use the report option on the page, or contact support.

Last updated: November 6, 2026