Content safety policy

1. Scope

This policy applies to work and episode titles, descriptions, covers, images, dialogue and subtitles in every language. The same standards apply to photorealistic and illustrated styles.

The following content may not be published. Labelling content as fictional or AI-generated does not make prohibited content permissible.

• Sexual exploitation or sexualization of children and young people, or sexual content involving minors
• Explicit sexual acts or sexually explicit content, and non-consensual sexual synthetic media
• Extremely graphic mutilation, gore or exploitative content involving people
• Content inciting hatred or encouraging violence against a person or group
• Specific instructions facilitating serious harm
• Disclosure of personal data without authority, or secrets such as passwords, authentication information and API keys
• Fraud, deceptive impersonation of real people, or fabricated photorealistic material used to mislead people about real events or people

2. Photorealism and real people

Photorealism is a visual style resembling photography and film. Do not present generated scenes as genuine recordings or evidence of real events, or falsely attribute statements or actions to real people.

When using identifiable elements such as a real person’s face, voice or name, confirm the necessary rights and consent. Creating sexual synthetic scenes without the person’s consent or impersonating someone to deceive readers is prohibited.

Provide context so readers can understand that a work is fictional. AI cannot guarantee it has independently verified a person’s identity, consent or the authenticity of material.

3. Review and holds

When automated review is enabled, images and text in all languages extracted from supported formats are reviewed in defined batches. Unreadable or missing material and uncertain safety, age or context are held for operator review rather than treated as passing.

Fictional conflict or a character’s age alone does not lead to an automatic ban. Context and the actual concern are considered together. Questionable material is held while an operator reviews the reasons and the parties’ explanations.

Passing automated review is not approval or a guarantee of copyright, legal compliance, identity or consent. It does not mean unseen material was reviewed or assign an all-ages rating.

4. Reports and responses

Use Reports and support to identify the work and provide verifiable reasons. File attachments are not supported. External reference URLs are recorded but not opened automatically. Do not submit passwords, authentication information, identity documents or other sensitive material.

A report does not automatically hide a work. An operator reviews the evidence and may suspend publication, issue a warning, restrict posting or dismiss the report. Material requiring preservation may be handled with restricted access.

Requesters and notified creators can view the reasons and history shared with them, and submit explanations or appeals. Another party’s original report and private operator notes are not disclosed.