Pairs with the commercial-use checklist for a job-by-job gate.

The three layers you must separate

Rights ownership and licensing
Rights ownership and licensing
LayerQuestion
Tool termsWhat does the generator’s license allow on your plan?
Underlying rightsAre you infringing someone else’s copyright, trademark, or publicity rights via the prompt or output?
Your dealWhat did you promise the client, employer, or platform in writing?

Passing one layer does not pass the others. A tool can “allow commercial use” while your prompt still targets a protected character. A client can “own the deliverables” while a platform bans undisclosed AI uploads.


Tool terms: read the boring PDF

Before monetized use, open the vendor’s terms for your current plan tier. Look for:

  • Commercial vs personal use
  • Ownership or license grant language for outputs
  • Restrictions on competing models, resale of raw outputs, or marketplace uploads
  • Voice / likeness features and consent rules
  • Whether terms change with little notice

Practice: save a PDF or screenshot of the terms version date in the project archive. When a dispute appears six months later, “I think it was fine” is not a paper trail.

Plan tiers matter. Free tiers often differ from Pro/Team/Enterprise.


Exact rules vary by country. Common themes:

  • Human authorship still matters in many systems for who can claim copyright in AI-assisted works.
  • Heavy human modification, selection, arrangement, and original elements can strengthen your position—but are not a magic shield for pure one-click outputs.
  • Copying protected works via prompts (“draw Mickey in…”) can create risk regardless of tool marketing.
  • Styles are treated differently from specific protected works; “in the style of a living artist” for commercial work is ethically fraught and sometimes legally contested—do not treat internet opinions as clearance.

If your business depends on exclusive ownership claims, get counsel—do not rely on a blog table.


Trademarks, brands, and publicity

High-risk prompt patterns:

  • Real logos and wordmarks for fake “partnerships”
  • Celebrity or influencer likeness for ads without deals
  • Competitor packaging that confuses consumers
  • Official uniforms, badges, or seals implying endorsement

Marketing teams get in trouble here faster than with abstract landscapes. When in doubt, invent fictional brands for comps—or use licensed brand assets under a real agreement.


Training-data and ethics (creator reputation)

Even when legal risk is ambiguous, reputation risk is not:

  • Clients ask how assets were made
  • Art communities police “style scraping” behavior
  • Marketplaces and contests add AI disclosure rules

A defensible process:

  1. Prefer original briefs and first-party references (your photos, your sketches).
  2. Avoid “make it exactly like [living artist’s famous series]” for commercial delivery.
  3. Disclose AI assistance when a client, platform, or brief requires it.
  4. Keep process notes so you can answer “how was this made?” without improvising.

See About this publication for how this site handles AI-assisted drafting of articles.


Client contracts: write the AI clause before the drama

Useful contract topics (examples—not templates):

TopicWhy
DisclosureWill you tell the client AI was used?
Tools listApproved tools / banned tools
OwnershipWhat transfers on payment?
IndemnityWho pays if a third party claims infringement?
RevisionsInfinite regen is not a revision policy
LikenessNo real-person cloning without releases
Stock + AICombined sources need combined licenses

If a client forbids AI entirely, that is a scope fact—not a moral debate during delivery week.


Platforms and marketplaces

YouTube, stock sites, app stores, game asset stores, and ad networks each invent rules. Before upload:

  • Disclosure toggles
  • Music Content ID collisions with AI tracks
  • Bans on AI-only assets
  • Advertising claim standards (especially for “photo of real product”)

Product accuracy matters: AI that invents features can become a consumer-protection problem, not only an art problem.


Music and voice special cases

From AI music and audio workflows:

  • Song generators may grant commercial rights only on paid plans
  • Voice cloning without consent is an ethics and legal minefield
  • Platform music policies can mute or claim videos regardless of your tool’s marketing

Archive license confirmations next to the WAV.


What to put in every project archive

Minimum:

  1. Brief
  2. Prompts + seeds/settings
  3. Tool name + plan tier + terms date note
  4. Final files
  5. Disclosure note (what you told the client)
  6. Sources of any non-AI assets (photos, fonts, stock)

This is the creative version of bookkeeping. Future you will need it.


International and employer twists

  • Employees: company IP policies often claim work product—including prompts and outputs
  • Agencies: client MSAs may pre-date AI; update them
  • Cross-border clients: governing law clauses matter
  • Open-source model weights: license of the model and of training data discussions differ from hosted SaaS

When in doubt, escalate early—not after the campaign launches.


Practical risk tiers (heuristic, not law)

TierExamplesDefault posture
LowerAbstract textures, invented objects, heavy human rebuildStill read tool terms
MediumCommercial product scenes, brand marketing, AI music bedsChecklist + archive
HigherReal people, famous IP, news events, political figuresCounsel + releases
StopNon-consensual intimate imagery, fraud, impersonation scamsDo not

How this site will talk about rights

  • We will not pretend a paragraph replaces an attorney
  • Scorecards mention “commercial posture” as a factor, not a warranty
  • Case studies are educational composites, not clearance certificates
  • When we discuss tools, verify current terms yourself—vendors change them

Published by Tabaconda LLC, Florida, USA. Not legal advice.

Useful physical tools for creator tools work. One optional product (see disclosure). Buy only what you need.

🎨 Back to Studio