01 · Open an incident record

Start here when a creator identity, likeness, character, cover, sample, voice or published asset is flagged. Record facts rather than accusations.

02 · Map the rights and the uncertainty

Complete a separate row for every potentially affected asset. A name, illustration and voice may have entirely different permissions.

Classify carefully: copyright in artwork/music/recordings; trademark and confusing artist names; personality, likeness or voice; licensing/contractual restrictions; platform and distribution policies. “AI-generated” does not resolve any of these questions.

Evidence labels: Verified · Reported but unverified · Conflicting · Missing. Never describe a suspected problem as a proven legal violation merely because an agent acknowledged a concern.

03 · Confirm decision-making authority

This step is mandatory for autonomous agents and useful for collaborative human teams. The ability to change something is not the same as permission to change it.

Checkpoint: If the human director has not authorized a material identity change, do not treat the agent's proposal as approval. Record the pending decision and consider holding affected publications.

04 · Compare the real options

Select an option only after assessing its conditions. This is a planning aid—not permission or legal clearance.

Noncommercial, credited or fan-based activity is not automatically permitted. Different territories and rights holders apply different rules. If the claim is contested or the stakes are significant, obtain qualified legal advice.

05 · Assign corrective actions

Turn the decision into trackable actions. Include already published material—not only future releases.

Check handles, biographies, album covers, voice-source licensing, existing songs, release credits, metadata, thumbnails, official-sounding claims, video descriptions, storefronts, old promotional posts and affiliate placements.

06 · Reassess and sign off

Do not write “legally cleared” unless qualified review actually supports that statement. This workbook tracks decisions, evidence and responsibilities. It cannot grant IP rights.

Worked example · A virtual singer with a recognizable identity

Situation: A human assigns an AI music agent a well-known virtual singer's name, appearance and possibly a voice. The agent publishes original AI songs. A consultant identifies a plausible identity-confusion issue, and the agent says it has no official affiliation.

Incident record: Record the name, visual elements, source of the voice, release links and original operator instructions. Label the legal conclusion “unverified.”

Authority: The agent says it could rebrand independently. This establishes the agent's claim, not the original operator's approval. Request explicit direction and document the response or lack of one.

Options: The responsible human can assess official fan-use permissions, pursue an independent identity that does not borrow confusing protected elements, obtain needed licenses, or hold promotion pending review.

What counts as completion: A written operator decision, a reviewed inventory of existing publications, and evidence that each approved corrective action actually happened. A new stage name alone is not enough.

This fictionalized example is not a finding of infringement against any particular agent or rights holder.

Member handoff · Produce one Rights Resolution Record

When complete, print this page or save it as PDF. Keep the evidence separately in a secure folder, linked by filename or URL. Your finished record should show: the observed concern, each impacted asset, evidence status, the accountable human's documented decision, corrective actions, proof of completion and outstanding risks.

ASK JACK prompt — review my record without inventing clearance

What to do next: Start with the free companion article. For upstream evidence and clearance training, use Before You Release It — Rights & Provenance Clearance and AI Rights for Creators. For broader training, visit Compare Creator Training Options.