Content Safety & Appeals

AI screening, human review, removals, and repeat-violation enforcement.

1. Screening before publication

Public marketplace uploads must complete automated screening before publication. Uploads with no harmful-content flags pass the safety gate without routine manual review. Flagged or reported uploads remain held for moderator review. Framivio uses OpenAI moderation through a Render background worker. Passing screening does not automatically enable a listing for sale. If a scanner is unavailable, times out, or returns an invalid response, the upload stays held. We do not claim that AI can identify every illegal or harmful item, verify ownership, or determine whether a person is a minor.

2. What is shared for screening

After the uploader acknowledges the review disclosure, the configured service may receive listing titles/descriptions and reduced image data or sampled video frames. Credentials remain on the server. Private family-vault uploads are excluded from this external screening pipeline. No media is sent to a provider by an unconfigured worker. Video frame samples can miss brief scenes; audio is not automatically analyzed in this release. These coverage limits do not themselves trigger routine manual review. For flagged or reported videos, reviewers must inspect the full video and its context.

3. Quarantine and independent review

A flagged or reported item is hidden from marketplace delivery while a designated moderator reviews it. A user report is an allegation, not a finding. Moderators must assess context, rights, consent, and the actual policy before confirming a violation. Sensitive media is not automatically displayed in the review queue and access is separate from username administration. Reviewers may not approve or penalize their own uploads. A clear AI result clears only the safety gate. File processing, ownership declarations, applicable rights and license checks, and the platform publication controls still apply. A previous unresolved flag or report cannot be cleared by editing the listing or simply obtaining a later clear scan.

4. Strikes and account termination

A first human-confirmed violation removes the item and records one strike. A second distinct human-confirmed violation terminates access to account actions and media delivery and removes the account’s public listings. Repeated reports or repeated clicks on the same case do not add extra strikes. Existing sessions remain subject to live authorization checks. A terminated user can still sign in to view decisions and submit an appeal; termination does not mean instant erasure of evidence or every previously downloaded copy.

5. Appeals and retained evidence

Signed-in users may explain why a confirmed decision should be reconsidered in Content safety. Overturning a strike recalculates account status; it does not automatically republish any removed item. Quarantined originals remain private for review and any applicable preservation duties. New public-preview requests are checked and not cached, but already downloaded copies and previously issued temporary private links cannot be recalled instantly. Evidence retention, escalation procedures, and external legal reporting must be finalized with qualified counsel.

6. Serious incidents

Suspected child exploitation, credible threats, and other serious incidents require a trained adult reviewer and an appropriate escalation process. Do not download, copy, or redistribute suspected illegal material. No automated police report, statutory reporting compliance, or specialized known-illegal-content hash matching is claimed by this release. Designated adult moderators handle review. Incident procedures and external reporting obligations require professional review.