You made a great image with an AI tool, and you want to put it on a product, a client project, or an ad. Reasonable question: are you allowed to? Most people go looking for the answer in copyright law. That’s the wrong drawer. The first thing that decides whether you can sell an AI image isn’t copyright at all — it’s a contract. Specifically, the license buried in the tool’s Terms of Service.
And the phrase everyone keys on — “commercial use” — isn’t a legal absolute. It’s a defined term that each company writes to mean whatever it wants. Two tools can use the same two words and grant you very different rights. Here’s how to read what you’re actually agreeing to.

License and copyright are two different things
Keep these separate in your head, because conflating them is where most creators get confused.
Track the cases yourself
Every current AI-copyright ruling — and the ones still being fought — lives in our AI Copyright Ruling Tracker. Filter by your creator type to see, in plain language, how each case affects your rights.
Copyright asks: does anyone own this image, and can they stop others from copying it? For purely AI-generated output, the answer under current U.S. Copyright Office guidance is often “no human owns it,” because a machine produced the expression.
The tool’s license asks something narrower and more immediate: has the company that ran the model given you permission to use the output, and for what? That permission lives in the Terms of Service you accepted when you signed up. It can grant you broad rights, restrict them to personal use, or hinge them on how much money you make.
So “can I use this commercially?” is really two questions stacked together: does my license permit it (contract), and can I stop anyone else from using the same image (copyright). This article is about the first one — the permission slip that actually governs your day-to-day right to sell.
What ‘commercial use’ actually means in the fine print
In plain terms, “commercial use” means using the image in connection with making money: selling prints or merch, using it in a paid client deliverable, putting it in advertising, or building it into a product you charge for. “Non-commercial” use means personal, educational, or experimental use where no revenue is attached.
The catch is that each license draws that line in its own place. Some grant commercial rights to every paying user automatically. Some reserve them for higher tiers. Some let you use the image but forbid reselling the image itself as a standalone product (a real distinction for print-on-demand and stock sellers). And nearly all of them carry use-based restrictions — no illegal, deceptive, or harmful content — that apply no matter what you paid. Reading the definition, not assuming it, is the whole job.

The subscription-tier trap (and revenue thresholds)
Here’s the mistake that quietly bites businesses: assuming that because you’re paying, you’re covered. Commercial rights are frequently gated by which plan you’re on, and sometimes by how much your business earns.
Midjourney is the classic example. Its terms have historically given paid subscribers ownership of the assets they create, but layered a revenue rule on top: companies above roughly $1 million USD in annual revenue are directed to a higher-priced Pro or Mega plan to use the service. Free-trial images, when trials existed, came with non-commercial strings attached. So a solo creator and a mid-size company can run identical prompts and sit under very different obligations.
The lesson generalizes: your commercial rights can depend on your tier and your size, not just your credit card. When your revenue grows, your license terms may need to grow with it — that’s a compliance checkpoint most people never revisit.
How the major tools handle it
A snapshot as of 2026 — treat it as a starting map, not gospel, because these terms change often:
- Midjourney — Paid subscribers generally own the images they create, subject to Midjourney’s own broad license back to the assets and the revenue-tier rule above. Commercial use is allowed on paid plans within those conditions.
- OpenAI (DALL·E / ChatGPT images) — OpenAI’s Terms of Use assign to you its right, title, and interest in the output, and permit commercial use. You may use images you generate for business purposes, subject to the usage policies.
- Stable Diffusion / Stability AI — Open models released under the CreativeML OpenRAIL-M license permit commercial use, but with a list of use-based restrictions you must pass downstream. Newer Stability models can require a paid membership for commercial use, especially above certain revenue — so “open” doesn’t automatically mean “free for business.”
- Adobe Firefly — Built deliberately for commercial safety, trained on licensed Adobe Stock and public-domain content, and marketed as safe for commercial output. Adobe also offers IP indemnification to eligible enterprise customers (more on that next).
Notice the pattern: the answer to “can I sell this?” is genuinely different across tools, and it hinges on the license text rather than on the image itself.
Indemnification: who pays if you get sued?
This is the part businesses underrate. A license can grant you the right to use an image and still leave you holding the bag if a third party claims the output infringes their work. Indemnification is the vendor’s promise to defend and cover you if that happens.
A handful of players offer forms of it — Adobe for eligible Firefly enterprise use, Microsoft and OpenAI for certain enterprise/Copilot customers, and stock-based generators like Getty and Shutterstock that trained on licensed libraries. But here’s the key qualifier: indemnification almost always attaches to business and enterprise tiers, comes with conditions, and rarely covers casual consumer accounts. If you’re running real commercial risk — a national ad campaign, a product launch — whether your tool stands behind you legally can matter more than image quality. For a hobby print, it’s largely irrelevant.
The ownership catch you can’t license your way out of
Now the sobering part. A license can give you permission to use an image. It cannot conjure a copyright that doesn’t exist. If your image is purely AI-generated with no meaningful human authorship, current U.S. guidance says it likely isn’t protectable — and no Terms of Service can change that, because the tool can only grant what it has.

The practical result: you might be fully licensed to sell an image and still unable to stop a competitor from using the exact same output, or something nearly identical, that the tool hands to someone else. You’ve bought a permission slip, not exclusivity. If exclusivity is what your business needs, meaningful human authorship — real editing, composition, original elements — is what starts to build a copyright the license alone never will.
What this means for you
You don’t need to be a lawyer to stay on the right side of an AI image license. You need a short, repeatable routine:
- Read the current terms before you sell. Find the tool’s “commercial use” and “ownership” sections. They’re usually short and change often — the version you read a year ago may be gone.
- Match your plan to your use and your revenue. Confirm your specific tier grants commercial rights, and check whether a revenue threshold pushes you to a higher plan as your business grows.
- Watch for resale limits. Using an image in your product is not the same permission as selling the image itself on stock or print-on-demand. Confirm the one you actually need.
- Prefer indemnified tools for real risk. For high-stakes commercial work, favor a service that offers indemnification on a business tier — and read what it actually covers.
- Keep a paper trail. Note which tool, which plan, and the date you generated each commercial image. If terms change later, you’ll want proof of what applied when you created it.
Bottom line: “commercial use” isn’t a law you can look up once — it’s a contract term each tool defines for itself, gated by tiers, revenue, and fine print, and separate from whether you own the image at all. Read the license, match your plan, and you can sell with confidence. Skip it, and you’re guessing with your business on the line.
This is general information, not legal advice, and AI-tool terms change frequently — always verify the current Terms of Service for the specific tool and plan you use, and for anything high-stakes or commercial, consult an intellectual-property attorney.
Sources & further reading:
- Midjourney — Terms of Service (ownership, subscription and commercial-use / revenue provisions)
- OpenAI — Terms of Use (ownership of Output)
- Stability AI — Licensing (CreativeML OpenRAIL-M and membership terms)
- Adobe — Firefly, generative AI and IP indemnification for enterprise
- U.S. Copyright Office — Copyright and Artificial Intelligence (human-authorship guidance)