Wonder & Worry
A guitar, brushes and a small canvas in a studio corner

A Shared Ethic for AI in art and music

Eight commitments for people who make things. By AI we mean generative tools that create new words, images, sound or video. “We” means anyone who chooses to work by this page.

People have lost work to these tools, and many had their work used without being asked. We name that first. We have also seen these tools help people make things they could never have made alone. Both are true, and we want to hold them together.

  1. Human first

    The work begins with a person... their intention, experience and feeling. Any AI we use helps carry it out. We do not use it to replace people we would otherwise hire.

  2. Consent

    We never clone a real person's voice or likeness, or train or prompt a tool to imitate a particular living artist, without permission. We never present work so people believe someone else made it. Every artist carries the influences they grew up with, and that kind of resemblance is welcome here.

  3. Credit and pay

    We support credit, licensing, collective bargaining and fair pay for the people whose work trained these systems.

  4. Say how it was made

    Wherever we share work publicly, we describe in our own words what we did and what AI did, and we name the tools we used.

  5. Keep the craft alive

    We keep practicing and learning without AI too, so our own skills keep growing.

  6. Purpose before profit

    This applies to all of us and asks the most of those with the most power: the companies that build and sell AI, then the labels, studios and platforms that use it, then individual artists. Everyone deserves a living. We share publicly only work we have reviewed and stand behind, and we never mass-upload, spam or deceive.

  7. Use energy with intention

    Everything these tools make uses electricity and water. We work in quick, rough drafts until a piece is worth finishing, make fewer throwaway versions, and save video, the most energy-hungry use, for when the work needs it.

  8. Respect

    We respect artists who refuse AI as much as artists who use it ethically, as these commitments describe, and we speak honestly about its risks and benefits.

What counts as imitation

Sounding like a singer you grew up with is resemblance, and resemblance is how music has always been passed on. The commitment on consent names three acts that aim at a particular person:

  • Cloning a real person's voice or likeness with a tool, without their permission.
  • Training or prompting a tool to imitate a particular living artist.
  • Passing off work so people believe someone else made it.

Influence, homage and genre stay welcome.

When in doubt

Would I be comfortable showing exactly how this was made to my audience, and to the artists whose work trained these tools?

A cross-section of soil in three layers

Choosing tools

Most AI tools were trained on work used without permission, so for now, using them falls short of full consent. We sort tools into three tiers by how their training data was gathered, and use the highest tier that can do the job.

  1. A

    Trained on licensed or owned work, backed by independent certification or a published list of sources.

    Our first choice.

  2. B

    The company says its data is licensed, and no one independent has confirmed it.

    We use these and say the licensing claim is unverified.

  3. C

    Trained on unlicensed or undisclosed data, or data challenged in unresolved lawsuits.

    We use these mainly for sketching, learning and private work. If we share the results publicly, we name the tool as Tier C.

Tools change tiers, so we recheck before relying on one. A tool that fits two tiers counts as the lower one.

We believe that, used this way, these tools can help more people create, heal, learn and share. We will revise this page as the law and the tools change, and we invite artists who are skeptical of AI to help shape it.

Print the one-page handout The evidence on art and music