Both MFI marks are issued under public standards. Anyone can read them, cite them, or audit a certification against them. The standards are versioned, dated, and published as PDFs.
The CSF certifies that a sound recording was produced without generative AI. Detection uses a documented scoring methodology with thresholds defined in the standard. Recordings flagged as indeterminate are reviewed by a trained analyst.
The standard defines the scoring algorithm, accepted detection sources, the thresholds for each tier (Verified, Confident, Probable, Attestation Required, Not Certified), the analyst review protocol, and the appeal process.
The CEP certifies that a sound recording was produced without grid correction, pitch correction, or generative AI. Foundational tracks must have been recorded simultaneously. Each submission is individually reviewed by a trained analyst.
The standard defines the production criteria, the analyst review checklist, accepted recording practices (multitrack, overdubbing, effects, mixing), prohibited practices (quantization, pitch correction, generative AI), evidence requirements, and the reconsideration process.
Standards are republished when methodology, thresholds, or review protocols change. Older versions remain available for reference.
| Date | Mark | Version | Summary |
|---|---|---|---|
| March 31, 2026 | CSF | v2.0 | Updated detection methodology, scoring thresholds, and analyst review protocols. PDF |
| March 31, 2026 | CEP | v1.2 | Refined the foundational-tracks criterion, added analyst checklist for hybrid live/studio recordings. PDF |
A short summary of how each certification flows from submission to public record. The full procedure is documented in each standard's PDF.
Each track is scanned with documented AI detection. Scores below the Verified threshold publish automatically. Scores in the indeterminate range are queued for analyst review. Scores above the Not Certified threshold are rejected with a written decision.
Every CEP submission is reviewed by a trained analyst against the published checklist. The analyst evaluates production methodology, not sound quality. Decisions are written and appealable within 30 days.
Certified recordings receive a unique ID and are published in the public registry. The audio file is deleted after review. A SHA-256 hash of the file is retained permanently for verification.