Is AI Art Ethical? Here's What I Think

I've been avoiding writing this post for months. Not because I don't have opinions. I have a lot of them. But because this topic has a way of making everyone angry no matter what you say, and I'm not really in the business of making everyone angry on purpose.

But people keep asking me what I think, and I keep giving these hedged non-answers that I don't actually believe. So here's what I actually believe.

My actual position

AI art is not inherently unethical to make or use. But the way most AI image models were built involved real ethical problems. And those two things can both be true.

I know that sounds like a dodge, but stay with me because I think the distinction actually matters a lot.

Making an image with an AI tool is not the same moral action as the decision to train a model on scraped unconsented data. The person generating a cool image of a robot dog is not the same as the executive who decided to build a dataset from billions of images without asking anyone. I think conflating those two things makes the conversation less useful, not more.

The part where I disagree with other AI enthusiasts

A lot of people who love these tools will tell you the training data issue is no big deal because "that's how all art works", you look at existing art and you learn from it. And I've said this myself in casual conversations. But I don't fully believe it anymore.

Here's the thing: when a human artist looks at other people's work and learns from it, they're not building a commercial product that can then reproduce stylistic elements at infinite scale on demand for paying customers. The commercial extraction is the part that's different.

When Adobe trains a model on licensed work and shares revenue with contributors, that's different from scraping millions of images from artists who never agreed to be part of a commercial dataset. Both result in an AI that can make images. But the relationship between the companies and the people whose work they used is completely different.

I think pretending otherwise is intellectually dishonest. And I see a lot of smart people I respect doing that because they like the tools and don't want to feel bad about using them. I get it. But I don't think it's the right call.

The part where I disagree with artists who hate all of it

Okay so I also push back on the other side of this conversation.

Some artists have taken the position that any AI-generated image is theft, that using these tools at all is participation in harm, and that anyone who generates images with AI is complicit in destroying the creative industry. And look, I understand the emotional weight behind that position. If my livelihood felt threatened by a technology I didn't consent to helping create, I'd be angry too.

But I don't think the all-or-nothing framing is accurate or useful.

A lot of people using AI image tools are not professional artists and were never going to hire one. Someone who uses Midjourney to generate a banner for their small Discord server was not a potential commission for a concept artist. The economic substitution claim is real for some use cases and pretty overstated for others.

Also, and this feels important to say: AI image tools have made visual creativity accessible to people who genuinely couldn't do it before. People with motor disabilities who can't draw. People without money to hire designers. People who have ideas but no technical ability to execute them. That's a real thing that has real value.

The technology is not purely extractive. Acting like it is, or like anyone who uses it is bad, ignores real benefits to real people.

What I think actually matters

I've landed on a few things I actually care about in practice, beyond the abstract debate.

Consent and compensation for training data matters and should be a bigger part of how we evaluate these companies. Adobe built a model with licensed content because they understood this would matter legally and ethically, not because they're saints, but because they anticipated accountability. Other companies have bet on accountability never coming. I think they're wrong, and I think choosing where to spend money based on this is a reasonable thing to do.

Using a living artist's name as a style prompt is something I personally don't do. I could be wrong that this is clearly a harm. I'm genuinely not 100% certain. But it feels to me like the most direct way an individual user can reduce negative impact on specific people. It costs me basically nothing. So I don't do it.

Disclosure matters when it matters. If I'm selling AI-generated art, or if an AI image is being used in a context where the audience might reasonably care, I think being honest about that is the right move. Not because AI images are lesser, sometimes they're genuinely gorgeous. But because audiences and clients deserve to know what they're getting.

The part that's actually complicated

Here's what I'm honestly not sure about: whether the harm from existing models is something that can be fixed through policy, or whether it's more like a genie that's out of the bottle in a way we can't fully walk back.

A lot of the proposals I've seen, licensing frameworks, opt-out registries, compensation pools, sound good in theory but seem really hard to implement in a way that actually reaches the artists who were affected. By the time any meaningful regulation exists, the models that were trained on scraped data will have been running for years and the companies that built them will have made their money.

I find that genuinely frustrating and I don't have a clean answer to it. I don't think anyone does right now.

Where I've landed

I'm going to keep using AI image tools. I'm going to keep being thoughtful about which ones I pay for and how I use them. I'm going to keep talking about the training data stuff because I think it matters even when it's inconvenient to people who like the tools, including me.

I don't think using these tools makes you a bad person. I also don't think the concerns about how they were built are overblown or are coming from people who just don't understand technology. Both camps in this debate spend a lot of time being condescending to the other side, and I think they're both missing something.

The technology is here. The question is what we do with it and who we hold accountable for how it was made. Those feel like the conversations worth having.