1. Scope

Moderation applies to any content uploaded by an artist (audio, covers, descriptions, credits, profile fields), to launch party rooms hosted on SynthCamp, and to messages exchanged through the on-platform reporting flow. SynthCamp does not host third-party comments or open chat outside launch parties.

2. What we moderate

• Copyright infringement: unauthorized use of someone else's recording, composition, or lyrics. Notified via /legal/dmca.
• Misleading AI declarations: mismatches between the artist's signed Creative Credits attestation and what our encoder reads from the audio file (embedded ID3 tags, declared AI-tool watermarks, JUMBF / C2PA presence). For example, a release declared fully human while the source carries a Suno encoder tag. AI use that is openly declared on a release is not a valid ground for a conformity report or moderation action. SynthCamp does not sanction AI-assisted or AI-generated music when the artist has correctly declared their tools and contributions in Creative Credits. Conformity reports filed on such releases under false_credits are auto-dismissed at submission. The moderation question is accuracy of the declaration, not the use of AI itself.
• Duplicate uploads:the same audio appearing under two different artist accounts, or the same audio appearing twice in the same artist's catalog. Detected by a four-pass dedup pipeline at upload time: SHA-256 byte match (intra-account, blocked synchronously), Chromaprint exact-string match, sliding-window Hamming distance over the per-frame Chromaprint (catches silence trims and light re-encodes that exact match misses), and ACRCloud Custom Recognition (catches pitch shifts, tempo changes, EQ-violent re-mixes). Cross-account hits are diverted to manual review.
• Illegal content (DSA art. 16): anything else that infringes EU or French law: hate speech, incitement to violence, CSAM, terrorist content, fraud, etc. Reportable via /legal/report. Child sexual abuse material discovered on the platform is removed from public view immediately and reported to the PHAROS portal at internet-signalement.gouv.fr for French law-enforcement referral and international coordination (PHAROS routes via IWF / INHOPE / NCMEC where relevant). The uploading account is suspended via the standard admin ban flow; the upload itself (encrypted on R2) and any associated transaction record on the `purchases` table are preserved as evidence and made available to authorities on judicial request. Connection identifiers (IP / user-agent / session timestamp) are retained for at most 12 months, as described on /legal/privacy and accessible to investigators within that window.
• Repeat infringers: accounts with three or more upheld notices within twelve months are subject to suspension under article 23 of the Digital Services Act. The moderation team operates a rolling 12-month counter per artist; when the threshold is reached the moderator issues a written warning and, after a reasonable period, a suspension.

3. How we decide

Three layers run on every upload. Each layer can divert a track to the pending-review queue; only a human moderator can take the terminal call.

• Algorithmic pre-flight (no decision): at finalize time, a SHA-256 lookup against the artist's existing fingerprints rejects a same-file re-upload with a structured 409. This is a deterministic check on bytes the artist already uploaded; we do not classify content.
• Encoder-side scan (automated decision, dispute pathway): our encoder reads ID3v2 / Vorbis / APE tags, runs a Chromaprint perceptual fingerprint, and probes for JUMBF / C2PA boxes. The output is persisted alongside the track for audit. When the scan disagrees with the artist's Creative Credits attestation, or the audio matches another committed track or a known commercial recording, the track is marked failedwith a clear reason in the artist's wizard banner. No human moderator intervention is required by default; the artist either deletes the track and uploads a different file, or clicks Request review if they believe the automated decision is wrong (next bullet).
• Human review (on artist request or admin escalation): when the artist clicks Request review on a failed track, it moves to pending_review and surfaces in the moderator queue. A SynthCamp moderator listens to the audio, reads the scan output and the artist's rebuttal, and approves (the track returns to encoding), rejects (the track stays failed with the moderator's reason recorded), or escalates. Bans are admin-only and are issued only after a documented pattern.

4. Notification of decisions (DSA art. 17)

When a track moves to pending-review, the artist sees a banner on the release page and on their catalog with the reason in plain language (for instance "ai_tool_metadata_detected:suno" or "duplicate_match:<trackId>"). They have seven calendar days to file a rebuttal from their dashboard (contradictoire). If no rebuttal is filed within that window the moderator decides on the available evidence; the artist remains free to re-upload an updated file at any point. Statements of reasons are also dispatched to the Commission's DSA Transparency Database.

5. Internal complaint-handling (DSA art. 20)

Any artist or notifier dissatisfied with a moderator decision can appeal it for free, in writing, by replying to the email that carried the decision or by writing to [email protected]. The appeal goes to a moderator who did not handle the original decision. We aim to complete the review without undue delay. The output is a reasoned written reply that supersedes or confirms the original decision, and is logged for the annual transparency report.

6. Out-of-court dispute settlement (DSA art. 21) and RGPD article 22

Beyond the internal complaint, EU users can refer the dispute to any out-of-court dispute settlement body certified under article 21 of the Digital Services Act once such bodies are designated for France. SynthCamp will cooperate in good faith with their procedure. Decisions taken with the help of automated processing carry an explicit human-review request button per article 22.3 of Regulation (EU) 2016/679 (GDPR), surfaced on the artist's pending-review banner.

7. Trusted flaggers (DSA art. 22)

SynthCamp will treat reports submitted by trusted flaggers, as designated by the Digital Services Coordinator of the country in which the trusted flagger is established, with priority and without delay. A trusted flagger should reach out to [email protected] with their certificate so we can register their identifier in our report-handling pipeline.

8. Misuse and bad-faith notices (DSA art. 23)

Repeated submission of manifestly unfounded notices, repeated uploads previously taken down, or other patterns of platform misuse may lead to a temporary or permanent suspension. The decision is taken by a moderator after a written warning where practicable. Suspensions are appealable through the procedure of section 5 above.

8 bis. No dark patterns (DSA art. 25)

SynthCamp commits to interface design that does not manipulate or deceive users. Concretely:

• Account creation, account deletion, consent withdrawal, and payment are equally accessible. None is hidden behind extra steps, dark visual hierarchy, or confirmation friction that the equivalent positive action does not have.
• No false urgency timers, no "X others are looking at this now" counters, no engagement-boosting badges. Live party countdowns are real-time clocks on actual scheduled events.
• No default-on opt-ins for data sharing with third parties, and no consent box where the pre-checked option is the consenting one. SynthCamp does not run a separate marketing opt-in: there is a single email-notifications toggle that gates both transactional mails (purchase receipts, party reminders for parties you joined, refund decisions, security alerts) and platform-activity mails (artist digest, weekly recap, new-follower notifications). That single toggle defaults to on and can be flipped off in your settings; we do not split it into separate transactional / marketing knobs and do not send any commercial promotional emails on top of platform-activity.
• No personalised feed and no editorial "staff picks" on the ranked surfaces. Every list is automated and labelled with the criterion that ordered it (see /legal/ranking).
• No nudges through visual hierarchy or button colouring toward a more expensive option. The pay-what-you-want selector offers a discrete numeric input alongside any preset, all rendered with identical visual weight.

A user who believes a SynthCamp interface element nudges them toward an unwanted action may report it via [email protected] and the design will be re-audited.

9. Statement of reasons retention

Every moderation decision is logged with a timestamp, the identifier of the moderator, the reason in plain language, and a snapshot of the evidence relied upon (scan readout, dedup match, notifier's message). Retention: ten years for published decisions, three years for unactioned reports, in line with our records-of-processing obligations under GDPR.

10. Annual transparency report (DSA art. 15)

SynthCamp publishes a yearly transparency report covering the previous calendar year. The report covers volumes of notices and orders received, decisions issued by the platform, complaints handled, accuracy of automated tooling, and any out-of-court dispute settlement outcomes. The report is published at /legal/transparency/moderation.

11. Single point of contact

Under articles 11 and 12 of the Digital Services Act, the electronic single point of contact for users, authorities, and the Commission is [email protected]. Communications are accepted in French and English.