Can You Use Midjourney, DALL·E, and Stable Diffusion Images Commercially in 2026?

TL;DR

• Midjourney: Paid plans grant commercial use; companies over $1M revenue need the Pro/Mega tier. No indemnification. All generations are public by default unless you're on Pro+ Stealth Mode.
• OpenAI image generation: OpenAI assigns users rights in outputs to the extent permitted by law, but commercial terms and indemnity depend on the product and account agreement. Check the current Terms and Service Terms before relying on it for high-stakes work.
• Stable Diffusion: The CreativeML Open RAIL-M license permits commercial use of the model, but the outputs come with no indemnity and no clearance against the LAION training data. Highest legal-risk option for enterprise.
• Adobe Firefly: One of the safest mainstream commercial tools. Trained on licensed Adobe Stock + public-domain content, with IP protections available on selected business/enterprise terms.
• Shutterstock AI / Getty AI: Designed for commercial use, indemnified, but generation quality lags the frontier tools.
• Google Imagen 3 / Veo: Commercial-use friendly via Vertex AI; enterprise indemnification available.

Want the full commercial workflow? Start with the Complete 2026 Guide to Using AI-Generated Images and Videos Commercially.

The terms of service for AI image tools change at least twice a year, and the differences between them matter — a lot. The wrong tool choice for a brand campaign isn't just an aesthetic issue; it's the difference between an asset you can defend and one that gets you a takedown six months after launch.

This article is a tool-by-tool breakdown of where each major image generator stands on commercial use as of May 2026. We cover the four questions that actually decide whether you can ship an AI image:

1. Permission: does the ToS allow commercial use on your tier?
2. Indemnification: if you get sued, does the tool maker pay your legal bills?
3. Training data: is the model trained on licensed or unlicensed data?
4. Output ownership: who owns (or has rights to) what you generate?

If any of those terms feel unfamiliar, read our Complete 2026 Guide to Using AI-Generated Images and Videos Commercially first.

Midjourney

Commercial use: Allowed on all paid plans (Basic, Standard, Pro, Mega). The Free Trial — when reopened — does not permit commercial use.

The $1M rule: Midjourney's terms state that if your business or your employer has annual gross revenue of more than $1,000,000 USD, you must be on the Pro plan (or higher) to use generated assets commercially. This catches a lot of marketing teams off-guard; the agency on the Standard plan generating assets for a Fortune 500 client is technically in violation.

Output ownership: Midjourney's terms grant you ownership of the assets you generate, subject to two big caveats:

• You grant Midjourney a perpetual license to your prompts and generated images.
• On non-Pro tiers, all of your generations are public in the community feed by default. Anyone — including competitors — can see and reuse them. Stealth mode (Pro+ only) is required for confidential work.

Indemnification: None. Midjourney explicitly disclaims responsibility for third-party-rights claims. If you get sued over a generated image, you're on your own.

Training data and lawsuits: Midjourney is named in the consolidated US artist class-action suit. The training set was scraped without licenses. This doesn't automatically mean your output is infringing, but it's a non-zero risk factor for enterprise legal teams.

When Midjourney is the right choice: Editorial work, concept art, hero images for small-to-mid businesses, agency ideation, internal mood boards. Avoid for: pharma/medical brands, anything depicting real public figures, anything that needs a clean indemnification chain.

OpenAI DALL·E and GPT image generation

OpenAI's image generation is now embedded across ChatGPT (DALL·E 3 / GPT-Image-1), the API, and Sora (video). The commercial terms differ by product.

ChatGPT (Plus, Team, Enterprise): Outputs can be used commercially. OpenAI's terms assign you ownership of the output to the extent permitted by law (recognizing that pure AI outputs may not be protectable — see Are AI-Generated Images Copyright Free?).

API: Per the OpenAI Terms of Use, you own the outputs and can use them commercially.

Indemnification: OpenAI's business and API terms may include an IP indemnity for eligible customers, with important exclusions. Do not assume a personal subscription has the same coverage as a business or API agreement; confirm the current contract before relying on it.

Training data: OpenAI has not published a full training-data list for every image model, and there are active lawsuits involving AI training practices. Contractual IP protections can reduce risk for eligible business/API customers, but they do not remove the need to review people, trademarks, and sensitive use cases.

When DALL·E / GPT image generation is the right choice: Enterprise teams that need a clear indemnification chain, integrated workflows inside ChatGPT (a lot of marketing teams already live there), and rapid iteration via prompt chat. Less ideal for high-end visual aesthetics — frontier-quality rivals Midjourney and Imagen still have an edge for hero shots.

Stable Diffusion (and the SDXL/SD3 family)

Stable Diffusion is open-weights — you can download and run it yourself, or use it through a hosted service like DreamStudio, Replicate, or Leonardo. The license is the CreativeML Open RAIL-M (for SDXL) or a newer Stability AI Community License (for SD3 family).

Commercial use of the model: Permitted under both licenses, with use-restrictions (no illegal content, no generating CSAM, no discrimination, etc.).

Output ownership: You own your outputs as far as Stability AI is concerned. But — and this is the catch — the underlying training data (LAION-5B and successors) was scraped from the web without licensing. Stability AI provides no indemnification for downstream outputs.

Indemnification: None from Stability AI for the open license. If you use Stable Diffusion through a hosted commercial service (e.g., Stability AI's own Enterprise plan or some Adobe integrations), check that specific contract.

The Getty Images v. Stability AI case: A multi-year UK lawsuit alleging that Stability scraped Getty's images. A ruling is expected by mid-2026. Whatever the outcome, it has elevated Stable Diffusion's perceived risk for enterprise legal teams.

When Stable Diffusion is the right choice: Internal R&D, fine-tuning on your own licensed data (this is a great use case — the open weights mean you can train it on your own brand assets), academic/research projects, low-stakes individual creative work. Avoid for: customer-facing brand assets that depict realistic humans or branded objects, unless you've built an additional clearance pipeline.

Adobe Firefly

Firefly is Adobe's answer to the enterprise commercial-use question, and it's the safest mainstream choice in 2026 for one reason: it's trained on licensed material.

Training data: Adobe Stock licensed library + public-domain content + Adobe-owned content. Adobe pays "Firefly Bonuses" to the contributors whose work was used in training, which has materially reduced contributor backlash.

Commercial use: Allowed on all paid tiers. The free tier has generation-credit limits but the commercial use grant itself applies.

Indemnification: Adobe offers Firefly IP protections on selected business and enterprise terms. This is one of the mainstream indemnification paths in the market, and it is a major reason regulated industries (pharma, finance, government contractors) consider Firefly.

Output ownership: You own your generated outputs. Adobe doesn't claim copyright. Outputs are tagged with C2PA Content Credentials so downstream verification (and the EU AI Act disclosure) is straightforward.

When Firefly is the right choice: Anything customer-facing for a regulated industry, anything where you need an indemnification chain, anything where you need C2PA provenance built in. Less ideal for the very latest aesthetic styles (Firefly is conservative by design) and not the strongest for photoreal humans.

Shutterstock AI and Getty Generative AI

Stock libraries that ran toward AI rather than away from it. Both built generators trained exclusively on their own licensed catalogs.

Commercial use: Yes — that's the whole point.

Indemnification: Yes. Both Shutterstock and Getty extend their existing stock-image indemnification framework to AI generations. Coverage limits depend on plan.

Quality: A step behind Midjourney/Imagen for pure aesthetic quality, but the differential is shrinking fast, and for many marketing tasks the quality is already more than sufficient.

Output ownership: Standard royalty-free license, similar to traditional stock. You get a license to use the image; the platform retains underlying rights.

When to choose: Conservative enterprise marketing, anything where the legal team already has a master service agreement with Shutterstock or Getty (this halves the negotiation time), anything where you want the appearance of stock photography rather than the aesthetic of AI art.

Google Imagen 3 / Imagen 4 (and Veo for video)

Google's image and video generators, accessible via the Vertex AI platform and the Gemini API.

Commercial use: Allowed via Vertex AI for both individual developers and enterprises.

Indemnification: Google offers a generative AI indemnification for outputs from approved foundation models on Vertex AI, including Imagen and Veo. The exact contract scope depends on the customer agreement; speak to your Google Cloud account team.

Training data: Google has not fully disclosed the training data. Google has stated it filters for known copyright issues and uses licensed and publicly available sources.

Quality: Imagen 3/4 produces some of the strongest photorealistic outputs available in 2026, particularly for human subjects and text rendering. Veo is among the leaders for short-form AI video.

When to choose: Teams already on Google Cloud, anyone needing the indemnification chain at enterprise scale, photoreal human imagery, video work where Veo's quality matters.

Quick comparison table

Practical decision framework

Pick Firefly if:

• You're in a regulated industry (pharma, finance, healthcare, government).
• You need an indemnification chain and a C2PA provenance trail.
• Your design team already lives in Creative Cloud.

Pick OpenAI if:

• Your team works inside ChatGPT all day and prompt-iteration speed matters more than peak aesthetic quality.
• You have a business or API agreement with explicit IP-protection language.

Pick Midjourney if:

• You're a mid-market marketing team or agency.
• You're under the $1M revenue threshold or on the Pro/Mega tier.
• Your work is editorial/concept-driven, not depicting real people or brands.

Pick Stable Diffusion if:

• You're doing R&D, fine-tuning, or controlled internal work.
• You have engineering resources to run it, and you've built your own clearance pipeline.

Pick Shutterstock / Getty AI if:

• You already have a master agreement with them.
• You want stock-style outputs with indemnification, and aesthetic-frontier quality isn't critical.

Pick Google Imagen / Veo if:

• Your stack is on GCP and you can get the Vertex AI indemnification baked into your existing contract.
• You need photoreal humans or short-form video at frontier quality.

Common mistakes (and how to avoid them)

1. Using Midjourney Standard at a large company. The $1M revenue threshold makes this a contract breach. Upgrade to Pro or budget for it.
2. Assuming Stable Diffusion outputs are safe because the license is open. Open ≠ indemnified. The open license covers your right to use the model; nobody is on the hook for your outputs.
3. Treating a personal subscription like an enterprise contract. IP protections vary by product and agreement; confirm your plan before relying on any indemnity.
4. Mixing Firefly outputs with non-Firefly outputs in the same composite and assuming the whole thing is indemnified. It isn't. Adobe's indemnification covers Firefly-generated material; the rest of the composite is your problem.
5. Forgetting that "commercial use allowed" doesn't override right of publicity or trademark law. No tool's ToS gives you the right to generate a photoreal celebrity endorsing your product.

How ImgIvy fits in

ImgIvy aggregates AI-generated images and videos from creators using multiple frontier tools, applies a uniform commercial-use license, and reviews for recognizable people and brands. We do not provide IP indemnification at the free tier — for that, you'll still want a Firefly or Vertex AI contract. What we do provide:

• A reliable pre-cleared source for ideation and mid-stakes campaigns.
• Provenance metadata where the upstream tool exposes it.
• A clear, written license you can hand to your legal team.

For high-stakes campaigns, layer ImgIvy with a tool that gives you contractual indemnification (Firefly, Shutterstock AI, Getty AI, or Vertex AI). The combination — broad sourcing through ImgIvy, indemnified generation for hero assets — is what most efficient marketing teams converge on in 2026.

Related guides

• The Complete 2026 Guide to Using AI-Generated Images and Videos Commercially
• Are AI-Generated Images Copyright Free? What the Law Actually Says in 2026

This article reflects the state of major AI image and video tools as of May 2026. Terms of service change frequently — always confirm against the live ToS before using a generated asset in production. This article is general information, not legal advice.