1. In short

• AI drafts, humans sign. Every episode passes human sign-offs before release.
• Every work carries a label, «صُنع بمساعدة الذكاء الاصطناعي» ("Made with AI"), and a "How this was made" card.
• We do not generate real people's likenesses or voices.
• We do not train AI models on your data.

2. How we make a work

AI assistants in the studio are advisory only. We may use assistants that score an episode's opening and ending, check cultural sensitivities, check facts, or compare a work with a sponsor's brief. What they produce is an opinion that an accountable person signs against. Nothing is published automatically.

3. The label and "How this was made"

3.1 Every published work carries the label «صُنع بمساعدة الذكاء الاصطناعي» ("Made with AI") on its page and in its player.

3.2 The "How this was made" card shows:

• the credits and roles,
• the human sign-offs the work passed (role and date),
• the sources for historical and heritage facts, when the work contains them.

3.3 When we post clips on social platforms, we mention AI use in the caption and turn on the platform's own AI-generated-content label. Sponsored works also carry the «برعاية» label (see the Sponsorship Disclosure Policy).

4. Real people and real voices

4.1 We do not generate the image or voice of a real person. Our characters are fictional.

4.2 Our celebrity partners appear only in real footage they recorded themselves. We only trim it, caption it, and adjust its sound and framing. Anything more needs a separate written consent from them.

4.3 Character voices are synthetic voices from a text-to-speech provider, styled for tone and dialect. We do not clone a real person's voice without their written consent.

5. AI you interact with

• "Talk to the character": a conversation with a language model playing a character from the work. The character does not reveal events beyond the last episode you watched. Replies are fiction and may be wrong. It is free within daily limits and not available in the Kids section. You can delete your conversations.
• Comment summaries: a language model may summarise how public comments on a work feel, in an aggregate report for the team and sponsors, without usernames.
• Safety checks: we scan our own works automatically with Azure AI Content Safety before release, in addition to human review.
• Ranking: editors choose what appears on the home page, with simple popularity signals. We do not build a profile of you to target you.

6. Providers

Models may change over time. Each provider has its own terms and acceptable-use policy, and we follow the strictest one.

7. What AI does not do here

• It does not publish an episode without human approval.
• It does not suspend or close your account.
• We do not use your data to train it.
• It does not make ads targeted at you.
• It does not talk to children.

8. Mistakes and corrections

AI can get things wrong, such as a distorted image or a wrong fact. If you notice an error, use the "Report" button or write to support@novellab.ai. We take down an episode with a serious error until it is fixed, and aim to do so within one hour of confirming the problem.

9. Principles and law

We follow the AI Ethics Principles of the Saudi Data & AI Authority, and we apply labelling and disclosure requirements where they apply to us, including platform rules.