# AI disclosure

## Roles

| Participant or system | Role |
| --- | --- |
| Dominique Devereaux | Project originator, source-lyrics author, creative direction, approvals |
| Charlie J | Prospective human collaborator; no contribution is claimed until accepted and delivered |
| ChatGPT / Codex | Research, fee calculation, document structuring, rights questions, repository implementation, and the September 13, 2026 corrective reconciliation |
| Claude and Gemini | Creator-reported intermediate tools during a model handoff; exact passage-level authorship is unresolved because no public generation log establishes it |
| Suno | Generator used for a pre-production audio demonstration |
| PLYR / AT Protocol | Listening and audience-response layer; not an authorship authority |

## What ChatGPT did

ChatGPT/Codex inspected the supplied project structure, organized requirements, compared public Fiverr pricing with a larger EP scope, calculated published platform-fee effects, drafted nonbinding business terms, identified clearance risks, and implemented this documentation.

## What ChatGPT did not do

It did not determine any person's human worth, obtain rights, sign an agreement, perform Charlie's contribution, verify every private asset, provide legal advice, or replace human approval.

## Suno demo terminology

Use **human-authored, AI-generated demo** when Dominique created the concept, lyrics, and narrative direction but Suno generated the audible performance and musical rendering.

Do not call the audio “human-generated” if the vocals, instrumental, melody, arrangement, or performance were generated by Suno. Do not call the composition entirely human-composed unless the human-created melody and musical structure can also be documented.

Preferred disclosure:

> Human-authored, AI-generated demo. Dominique Devereaux created the concept, source lyrics, and narrative direction. Suno generated the audible demo performance and production. This is a disclosed pre-production experiment, not a human vocal performance, finished commercial master, or autonomous AI authorship claim.

## Consent boundary

No model may be trained or fine-tuned on a collaborator's voice, likeness, stems, sessions, or unreleased work without separate affirmative permission. AI use outside the approved scope must be disclosed before delivery.

## Model-handoff and correction disclosure

On September 13, 2026, Codex misread the project as only an EP archive and proposed removing intentional series-development and educational-resource functions. PR #25 was closed without merge, so it did not change the live site. The incident is documented in `docs/incident-2026-09-13-site-function-audit.md` and now serves as an acceptance test in `LLM-REVIEW-PROTOCOL.md`.

Future model attribution must be based on a preserved generation record. GitHub account ownership, commit timing, or writing style alone does not verify which model produced text.
