How to use DALL-E Ethically in your Graphic Design Work

Updated March 2026.

DALL-E is a model developed by OpenAI that generates images from text descriptions. You write what you want; it produces an image. The technology is genuinely impressive — and it raises real questions that every designer using it professionally should be able to answer.

The short version: yes, you can use AI-generated images in your work. But there are conditions worth understanding before you do, and the ethical questions go beyond what any terms-of-service document covers.

This post is not about whether to use AI image tools. That decision is yours, and the case for using them thoughtfully is strong. It is about how to use them in a way that is honest, professional, and defensible.

Copyright in AI-generated images is unsettled law in most countries. The basic principle in copyright law is that there must be human authorship for a work to attract copyright protection. An image generated entirely by an AI model, with no human authorship beyond the prompt, currently has a weak claim to copyright in most jurisdictions — including the UK, Australia, and the US.

What this means in practice: you may have usage rights granted by the platform’s terms of service, but you may not own copyright over the image in the traditional sense. For most commercial design projects, the distinction matters.

DALL-E: OpenAI’s current terms allow commercial use of generated images. You can use them in client projects, merchandise, and publications. You cannot, however, claim copyright over the output in most legal contexts.

Adobe Firefly: Built specifically with commercial use in mind, Firefly is trained on licensed and public domain imagery. Adobe provides indemnification for commercial use on paid plans — meaning if a client is concerned about IP risk, Firefly is currently the most defensible option.

Midjourney: Commercial rights are available on paid plans. On free plans, output is licensed under Creative Commons (non-commercial). If you are using Midjourney in client work, ensure you are on the appropriate plan and have read the current terms.

The landscape here changes. Check the current terms of whatever tool you are using before delivering AI-generated assets to a client.

This is where many designers are uncertain, and the uncertainty is worth addressing directly.

There is no universal requirement to disclose the use of AI tools — in the same way there is no requirement to disclose that you used a stock image library or a Photoshop filter. The question is whether the client would reasonably expect to know.

If a client has commissioned original illustration and you deliver AI-generated images styled to look like illustration, that is a misrepresentation. If a client has commissioned a logo and you use AI to generate initial concepts that you then develop and refine, that is a workflow choice — and one most clients would be comfortable with if you explained it.

The practical guidance: set the expectation early. Tell clients what role AI tools play in your process. Most clients are not opposed to AI-assisted work; they are opposed to being surprised by it after the fact. Transparency protects both of you.

This argument comes up regularly, and it deserves a considered response rather than a dismissal.

AI image tools can replicate certain types of visual output — stock-style photography, illustration in recognisable styles, generic brand imagery — quickly and cheaply. For designers whose work sits at that end of the spectrum, the competitive pressure is real.

What they cannot replicate is design thinking. They cannot understand a brief, challenge a client’s assumptions, develop a visual language that is specific to a brand, or make the hundred small decisions that distinguish good design from passable output. The designers who will be most affected by AI tools are not the ones with the strongest thinking — they are the ones who have not developed it.

This is not a comfortable argument, but it is an honest one. The answer is not to avoid AI tools; it is to develop the foundations that make them a weapon in your hands rather than a threat to your work.

It is useful to think concretely about the kinds of projects where AI image generation is genuinely appropriate, and where you should be more cautious.

Where it works well:

• Concept development and mood boarding. Using AI to explore visual directions early in a project, before committing to photography or illustration, is one of the clearest use cases. The output is never final; it is a communication tool for client conversations.

• Placeholder imagery for layout and presentation. If you need realistic placeholder content to show how a layout will read, AI generation is faster and more contextually appropriate than stock photography.

• Texture and background elements. AI-generated textures, patterns, and abstract backgrounds can be used as design elements where the question of authorship and copyright matters less.

• Personal and student work. Building your prompt vocabulary and learning what these tools can do is valuable. Your own projects are the right place to do that experimentation.

Where you should be cautious:

• Representing AI output as original photography or commissioned illustration. This is a misrepresentation.

• Generating images in the style of a living artist without disclosure. The ethical and legal status of this is genuinely contested, and the design community has strong feelings about it.

• Using AI imagery where cultural sensitivity applies. The models can produce outputs that reinforce stereotypes or that are culturally inappropriate in ways that are not always obvious. Apply the same critical eye you would to any imagery you source.

• High-stakes identity projects where IP ownership is essential. For a brand that will need to defend its visual identity, the copyright uncertainties around AI-generated elements are a real risk.

One of the best early examples of AI image generation used well in professional design is the &Walsh rebrand for Isodope. The team used DALL-E as part of their creative process, not as a replacement for it. The AI-generated outputs informed the visual direction; the final work was the result of design thinking applied to and beyond those outputs.

The itsnicethat writeup is worth reading — not just for the work, but for how the team talked about the process. They used it as a tool with a clear intention. The result is design work, not AI output.

That is the model. Use the tools with intention. Know what you are asking for. Apply your judgement to what you get.

If you want practical guidance on getting better outputs from AI image tools, read our post on getting started with DALL-E and AI image generation. For the broader question of AI language models in your workflow, see how to use ChatGPT ethically in your design work.

The foundations that make all of this coherent — design thinking, professional practice, client communication — are what we teach at The Graphic Design School. Our Certificate IV in Graphic Design covers both the craft and the context.