AI Content Moderation
How AI works quietly behind the community moderation workflow — flagging risk in the background, while people keep the final call.
MyLimb is a community-based platform where members create posts, comments, and magazine-style articles. That activity is the heart of the product, but it also creates a moderation challenge: new and edited content needs to be checked without slowing people down or removing human judgment from sensitive decisions.
Thoughtful Robots helped MyLimb use AI as a first-pass safety checker. The system scans meaningful content changes in the background, sends text and image URLs to OpenAI's moderation service through RubyLLM, saves the moderation result, and creates a report when something looks potentially unsafe.
The AI does not directly delete content. It feeds risky items into the existing moderation workflow so a human moderator can review, dismiss, resolve, or remove them.
Technical Foundation
The implementation uses a reusable Rails concern called AiModeratable, persistent AiModerationRecord entries, asynchronous jobs, OpenAI moderation through RubyLLM, and MyLimb's existing report workflow.
Models opt into moderation by defining the text and images that should be reviewed and the fields that should trigger a new moderation pass. This keeps the logic reusable across community posts, community comments, and magazine articles.
Step-By-Step Flow
- A user creates or edits content, such as a post, comment, or magazine article.
- The system checks whether AI moderation is enabled for that account.
- The system checks whether meaningful content changed.
- The content is queued for background review so the user does not wait.
- The backend prepares the relevant text and image URLs for AI.
- OpenAI moderation is called through RubyLLM.
- AI returns safe or potentially unsafe signals across moderation categories.
- A moderation record is saved with the checked content, image URLs, result, flags, and timestamp.
- Safe content is marked approved.
- Flagged content creates an automatic moderation report.
- A human moderator reviews the report in the normal workflow.
- If the content is edited again, older AI checks are marked outdated.
Short Summary
Thoughtful Robots helped MyLimb add AI moderation as a background safety layer for community content. The system only runs when meaningful post, comment, article, or image content changes; sends the prepared content to OpenAI moderation through RubyLLM; stores an audit record; approves safe content; creates reports for flagged content; handles stale checks; and keeps final decisions with human moderators.