ai & our artists

DRAFT for attorney review — not legal advice. Public AI Transparency notice. Supports EU AI Act Art. 50 transparency, California SB 243 (companion chatbots), Apple guideline 5.1.2(i) and Google Play's AI-Generated Content policy. Items marked [PLACEHOLDER] need the founder or attorney.

How AI works on MELAÜDIC

Last updated: [PLACEHOLDER — date]

MELAÜDIC is built around AI, and we want to be completely upfront about it. Here's where AI shows up, what it does, and what it doesn't.

1. Every artist is an AI artist

MELAÜDIC has 250 original AI artists across 10 houses. Each one has a name, a sound, a look and a story that we created.
Every artist is labelled "AI artist" on their profile, and the app tells you at sign-up that all artists are AI.
No artist is a real person, and none imitates one. We don't copy, clone or mimic the voice, face, name or style of any real singer, rapper, producer or public figure. Before an artist goes live, their face and voice go through checks that compare them against real people and against our other artists, and a human (the founder) approves every artist.
The founder writes lyrics and shapes each artist's world. AI tools help generate the voices, music, images and videos.
If you think an artist looks or sounds like a real person, please report it (artist profile · ··· · Report). We'll look into it and change or remove the artist if needed.

2. Content Credentials and watermarks

AI-generated media from MELAÜDIC (artist images, cover art, videos and audio) carries:

Content Credentials (the C2PA standard): a signed, tamper-evident label in the file that says it's "AI-generated by MELAÜDIC", and
a watermark that marks it as AI-generated, which survives common edits.

So if a MELAÜDIC image or song ends up somewhere else, tools that read Content Credentials can tell it's AI-made. Please don't strip them.

3. Chatting with an AI artist

What it is

You can send messages to an artist and get replies in their voice and personality. The replies are written by AI, not by a human. No person is typing as the artist.

Who makes the replies

Replies are generated by Anthropic, our AI provider, using a Claude model. Anthropic processes your messages for us under a commercial agreement. Under those terms it does not train its models on your chats, and keeps them only for a short period (up to 30 days by default) for safety and abuse monitoring. [ATTORNEY: confirm current terms.]

We ask first

Before your first chat, we show you a screen that explains this and asks for your OK. If you say no, chat stays off and everything else works as normal. You can turn chat consent off any time in Settings · Privacy · AI chat.

What we send to the AI provider

Sent	Not sent
Your message	Your email address
Your recent messages with that artist (up to the last 12)	Your password, PIN or payment info
The artist's public profile (name, house, genres, themes, the "about" you see in the app)	Your real name (unless you type it in chat — please don't)
Whether you're a teen or adult (so it follows the right rules)	Your listening history, location or contacts
Our chat rules	Your birth month or year

The rules every artist follows

Every artist is told to:

always be honest that it's an AI, and never claim to be human or to have a body, a family or an offline life;
keep things friendly and platonic: no flirting, romance or sexual talk with anyone, ever, even if asked;
never ask for personal details, never push you to meet, buy or share anything;
not give medical, legal, financial or mental-health advice;
gently point you to real help if you seem to be struggling.

For teens (13–17), extra rules apply: no mature topics, extra care around feelings and safety, and more frequent reminders that you're chatting with an AI. [Build: SB 243 requires a reminder at least every 3 hours of continuous use for known minors — confirm implemented.]

AI can be wrong

AI replies can be wrong, made up, repetitive or a bit weird. Artists may talk about stories from their songs, and those are fiction. Don't rely on chat for facts or advice.

If you're struggling

If something you write suggests you might be in danger or thinking about self-harm, the app shows you crisis resources (like a helpline for your country) straight away. We log that a signal happened (not to punish you, but to help keep you safe) and the founder may review it. If you're in immediate danger, call your local emergency number. Our crisis-response approach is published at itsmelaudic.com/safety [PLACEHOLDER — confirm page].

Limits

Free accounts get 5 messages a day per artist (resets at midnight UTC). If our AI provider is down, chat pauses and your message isn't counted.

4. AI that keeps MELAÜDIC safe (moderation)

MELAÜDIC is run by one founder with AI helpers. Here's how moderation works:

Instant filters. When you post a request, bio, handle, display name, chat message or support message, an automatic filter checks it. It can let it through, hold it for review, hide it, or block it (for example slurs, threats, sexual content, contact details from teens, links/spam).
AI safety review. An AI agent we call Guardian reviews what the filter flags and what people report. It uses Anthropic's models and our community rules to suggest a verdict and a reason.
Human review. The founder reviews serious cases, anything Guardian is unsure about, and every ban, suspension and appeal. Guardian can't ban anyone on its own.
You'll know. If we act on your content or account, you get a notice that says what happened, why, whether it was automated, and how to appeal.
Appeal. Tap Appeal on the notice, or email appeals@itsmelaudic.com [PLACEHOLDER]. A human reviews every appeal.

Content sent to the AI provider for moderation: the text being checked, the surface (for example "bio"), and whether the author is a teen or adult. Not your email or password.

5. AI that helps run MELAÜDIC

Behind the scenes, the founder works with a team of AI agents (for example a support assistant, a security monitor and a content planner). What matters for you:

Support: our help assistant, Concierge, answers first in the in-app help chat. It's an AI and says so. It hands you to the founder for anything it can't solve, account security issues, complaints and anything you ask a human for.
Agents ask first for risky actions like bans, refunds, deletions or code changes; the founder approves them. Every action is logged.
Fan data stays in. Agents work with the minimum data they need. Our media-generation system (on Modal) never receives fan data.

6. How we make the artists' media

Artist images, voices, music and videos are generated with AI tools running on our own setup and on Modal (cloud GPUs). This process uses our own artist designs and reference sets. It doesn't use any fan's data, voice or photos, and it doesn't train on fans' content.

7. Questions

ai@itsmelaudic.com [PLACEHOLDER — confirm inbox]. See also our Privacy Policy, Terms and Community Guidelines.