Skip to content

Use AI with care

How to make clearer ethical choices about AI images

Check training data, source permissions, artist credit and client expectations before using an AI image tool.

The quick answer

There is no single label that settles whether every use of an AI image tool is ethical. Look at how the model was trained, what you upload, who the image depicts and how the result will be used.

Set clear limits for your own work. Use sources you have permission to use, check provider claims and be honest about important edits.

In this article
  1. Name the concern before choosing a tool
  2. Check the exact model’s training policy
  3. Use public summaries without overreading them
  4. Make careful choices with references
  5. Consider the people affected by the image
  6. Turn your values into a project note
  7. Sources

Keep these in mind

  • Training data, your uploads and the final image need separate checks.
  • A provider’s policy is evidence of its stated approach, not proof of every outcome.
  • Agree consent, credit and acceptable changes before the job.

Name the concern before choosing a tool

When someone asks whether AI steals art, they may mean several things: unapproved training, copying a specific picture, imitating an artist or replacing paid work. Those concerns need different answers.

Write down what matters for your project. You might require documented training sources, permission for reference images, no imitation of a named living artist, and a clear description of generated parts. These are useful working rules even when people disagree about AI more broadly.

Check the exact model’s training policy

Look for named data sources, licensing information and an explanation of how creators are involved. “Responsible AI” on a sales page is not enough detail to judge those questions.

Adobe says Firefly was trained on licensed content, including Adobe Stock, and public-domain content whose copyright has expired. That is Adobe’s stated approach for Firefly. It does not cover every partner model in an Adobe app or every other service.

Also ask whether your own uploads may be used for future training. The sources used to build a model and the treatment of new customer files are different questions.

Read more: Adobe Firefly FAQ: Training data and content use

Use public summaries without overreading them

The European Commission provides a form for training-data summaries. It covers providers of general-purpose AI: models used for many tasks. There is a transition period for models available before August 2025. The summaries help you check data sources. They do not certify every result as lawful or ethical.

Read the scope and model version. A summary is also not a complete permission record for every image you might generate. If a key answer is missing, record that gap rather than filling it with a guess.

Read more: European Commission: Public summaries of AI training content

Make careful choices with references

Use your own approved work or references whose terms allow the planned use. Do not assume an image found online is available for any purpose. Keep the creator, source and permission beside the file.

Describe the visual qualities you need: rough paper, flat colors, wide shapes or soft side light. This can communicate an idea without asking the tool to imitate a particular artist. When a client wants an artist’s distinct contribution, discuss commissioning that artist.

Credit and payment are separate decisions from tool choice. Make them explicit when you work with illustrators, photographers or other contributors.

Consider the people affected by the image

Check whether an edit changes someone’s identity, body, age or role in a misleading way. Ask for approval of sensitive portrait changes. Review whether repeated image choices reduce people to stereotypes.

Think about what viewers will believe. A fictional scene for a concept board serves a different purpose from a picture offered as evidence of a real event. Explain meaningful invented elements in words the audience can understand.

Turn your values into a project note

A short note makes your choices reviewable. Revisit it when the client, model or purpose changes.

  1. List approved tools and the evidence behind your choice.
  2. List permitted references and uses of people’s likenesses.
  3. Write down what may change and what must remain accurate.
  4. Agree credit, client approval and how generated parts will be described.
  5. Keep the final decision with a person who can check the brief.

Sources & further reading

Product features can change. Check the linked documentation for the version and service you use.