An independent certification mark for sound recordings produced without generative AI. Awarded under a published scoring methodology with analyst review for ambiguous results, operated as a broader public registry mark.
Certified Significant Form certifies a sound recording produced without generative AI. The mark is structural, not stylistic. It does not evaluate genre, lyrical content, production aesthetic, or commercial merit. It certifies one structural fact about the recording: that its musical content was not generated by AI systems.
The criterion covers whole-cloth prompt-to-music output, voice cloning of vocal performances, and AI generation of any element that would otherwise constitute a performed musical contribution. AI-assisted production tools that do not originate creative content, such as mastering assistants, noise reduction, and similar utilities, are not within scope. The full definition is in the published standard.
Each recording is scanned with a documented detection methodology. The scan returns a probability score indicating the likelihood that the recording contains AI-generated content. The score is mapped to one of five tiers defined in the standard. Recordings in the verified and confident tiers are certified automatically. Recordings flagged as attestation-required or higher are routed to analyst review.
| Tier | AI probability range | Outcome |
|---|---|---|
| Verified | 0% – 35% | Automatically certified |
| Confident | 35.01% – 50% | Automatically certified |
| Probable | 50.01% – 65% | Analyst review |
| Attestation Required | 65.01% – 80% | Artist attestation required, then analyst review |
| Not Certified | above 80% | Rejected with written decision |
Tier thresholds, the detection methodology, and the analyst review procedure are defined in the public standard. Older versions remain available for reference.
Certified recordings receive a unique CSF identifier and are entered into the public registry at registry.musicalform.org. Each entry is linked to the recording’s ISRC, the certification date, the tier, and the artist and label metadata. Any third party can verify a recording’s status against the registry.
CSF is issued to a specific sound recording. It applies to the recording, not to the artist, label, or performer. The mark is intended for recording artists, independent labels, distributors, digital service providers, and listeners who want a verifiable structural claim that a recording was not generated by AI.
The Institute issues two certification marks. CSF is the broader public mark and certifies only that a recording was produced without generative AI. The Certificate of Embodied Production (CEP) is the stricter mark and adds the no-grid-correction and no-pitch-correction criteria. A recording certified under CEP necessarily satisfies the CSF criteria; the inverse is not true.
The full criteria, tier thresholds, detection methodology, and analyst review procedure are defined in the public standard document. The standard is versioned, dated, and published as a PDF. Older versions remain available for reference.