An editorial and operational checklist for safely producing AI images, audio, and video while protecting inputs, identities, licenses, reviews, and provenance.
Record the origin of every ingredient
References, photographs, music, voices, fonts, logos, scripts, and internal datasets must enter the workflow with a known source and permission. Visible inspiration does not mean permission to copy. The team distinguishes owned, licensed, public, generated, and client-supplied material, along with the restrictions on each use. If provenance cannot be explained, the asset should not proceed to high-impact commercial production.
Responsible synthetic production begins before generation. It requires checking input rights and consent, provider terms, represented identities, claims, final use and territory. It then preserves provenance, decisions and versions. Legal review depends on country and case; no editorial checklist replaces specialist advice.
Review provider terms for the specific use
Conditions can change by plan, model, country, and material type. Before promising exclusivity or broad rights, the provider's statements on inputs, outputs, retention, commercial use, and claims should be reviewed. The organization should retain the version of the terms applicable to the project and escalate questions to legal counsel. There is no universal answer about ownership of all AI-generated content.
A person's identity requires a higher standard
Faces, voices, bodies, signatures and other recognizable traits may involve consent, contracts, reputation and regulation. Permission for traditional photography must not be assumed to allow a replica or unlimited variations. The brief must specify the person, use, territory, media, duration, permitted editing and withdrawal process. For real people or sensitive contexts, specialist review happens before generation, not only before publication.
Preserve provenance and transformations
The production record can document the tool, date, operator, sources, relevant instructions, edits, and approved version. Standards such as C2PA provide mechanisms for attaching provenance credentials to compatible content, but technical implementation does not replace internal documentation or prove a claim is true on its own. Provenance helps reconstruct the asset's history and respond when a platform, customer, or audience asks how it was created.
Approve according to channel, context and consequence
An internal exploration, an organic post, and a mass campaign do not require the same level of control. Final review checks rights, unwanted similarities, anatomy and continuity, claims, cultural sensitivity, disclosure where appropriate, and contract compliance. It also retains a contact who can quickly remove or correct material. Production speed creates value only if the organization can explain and defend what it publishes.
