New Jeanland

Having Ethical Concerns About Generative AI Does Not Make You the Internet’s Compliance Officer

Anti-generative-AI culture has developed the same reputation stereotypically associated with militant veganism: a tendency to blur the line between a personal ethical choice and a universal moral obligation, coupled with an almost compulsive need to announce that position in conversations where nobody asked.

What makes the 2026 version especially irritating is how often that moral certainty is paired with a shallow understanding of the technology being condemned.

The Vegans Have Been Dethroned

For years, the cultural stereotype was simple: How do you know someone is vegan? Don’t worry. They’ll tell you.

The vegan is no longer literally a vegan.

Now, post a meme. Change your avatar. Mention ChatGPT. Use an AI-generated dog as a background image. Somewhere, someone is preparing to explain data centers to you. They may also explain copyright law, water consumption, labor exploitation, artistic integrity, and the impending death of human creativity, regardless of whether any of those subjects were actually under discussion.

Sometimes, you do not even have to be using generative AI. Suspicion is enough. A digital illustration looks slightly too polished. A sentence sounds unusually formal. Someone notices an odd finger in the background of an image. You use an em dash. Suddenly the comments section has convened an ethics tribunal.

The problem is not that people have ethical objections to AI. There are plenty worth discussing. The problem is the assumption that having those objections automatically grants you jurisdiction over every use of the technology you encounter.

“AI” Is Not an Ethical Category

Generative AI raises serious questions about training data, licensing, compensation, copyright, labor displacement, synthetic media, impersonation, misinformation, environmental cost, and commercial substitution. Those questions are worth having precisely because they are not all the same question. “AI is unethical” compresses them into a single moral category and then treats the category itself as the argument.

Nobody talks about “using the internet” as though reading Wikipedia, watching Netflix, running a phishing operation, and operating a botnet are morally interchangeable merely because all four require computers connected to a network. Yet “you used AI” increasingly functions as though it tells us everything we need to know. It does not.

Using a language model to brainstorm article titles is not the same activity as fabricating evidence. Generating a joke image for a Discord server is not the same thing as producing nonconsensual sexual images of a real person. Asking a model to reorganize notes you already wrote is not ethically identical to deliberately replacing an illustrator with a system instructed to imitate that illustrator’s work.

Some uses are relatively easy to defend. Some are genuinely difficult. Some are plainly abusive. The relevant question is not merely whether AI was involved. It is what the technology was used to do, under what circumstances, and with what consequences. Once that distinction disappears, “AI” stops being a description and becomes a contamination label.

The Confidence-to-Knowledge Ratio Is Catastrophic

That flattening would be irritating enough on its own. It becomes worse when the moral certainty rests on technical claims that are either wrong or badly oversimplified. You hear that image generators simply copy and paste fragments of existing artwork. That every generated image is automatically copyright infringement. That every prompt carries some enormous environmental cost. That because copyrighted material may have appeared in a training dataset, every downstream use of the resulting model is theft. These claims often gesture toward real controversies. That does not make them accurate descriptions of those controversies.

There is a difference between arguing that copyrighted material was acquired improperly for training and claiming that a model retrieves and pastes copyrighted images whenever it generates an output. There is a difference between plagiarism, memorization, stylistic imitation, copyright infringement, and dataset provenance. There is a difference between the cost of training a large model and the marginal cost of an individual inference. Those distinctions are not pedantic. Ethical reasoning depends on factual premises. If the factual premise is wrong, moral intensity does not rescue the argument.

This is hardly unique to AI. People routinely learn enough vocabulary about a complicated subject to participate in the discourse without learning enough to understand where one concept ends and another begins. Generative AI has simply become an unusually efficient machine for producing ten minutes of terminology followed by several months of confidence.

The Artistic-Theft Argument Gets Murky Fast

The same problem appears in discussions of art, where several distinct objections are often collapsed into the language of “theft.”

One argument concerns training data: whether copyrighted work should be copied into datasets without permission, compensation, or licensing.

Another concerns outputs that closely imitate a living artist’s recognizable style.

A third goes further and treats aesthetic influence itself as a form of exploitation.

Those are not the same claim. The first two raise difficult questions about scale, consent, commercial substitution, and the acquisition of source material. The third begins to imply a kind of ownership over influence that we have never generally granted human creators against one another.

Human art has always developed through exposure, imitation, reference, genre, convention, adaptation, parody, homage, and creative borrowing. We do not ordinarily give painters exclusive ownership over techniques, writers ownership over genres, or musicians ownership over every aesthetic feature another person might absorb from their work. That does not make machine training socially identical to human learning. It plainly is not. A company ingesting millions of works is not the same event as a student spending an afternoon in a museum.

That difference has to be explained rather than merely asserted.

Scale may matter. Automation may matter. Commercial displacement may matter. Dataset acquisition may matter. “Machines are different” is not yet an argument about how, and “style theft” becomes increasingly slippery when it begins implying rights over mood, composition, technique, or influence that copyright law and artistic culture have historically resisted granting even against other humans.

From Ethics to Purity

The most interesting shift in anti-AI culture is not the existence of ethical objections. It is what happens when those objections harden into a purity test.

“I don’t use generative AI” is a personal choice.

“I avoid it because I think there are ethical problems with it” is an ethical judgment.

“I think certain uses are harmful and should be discouraged” is a normative argument.

All perfectly coherent positions.

The trouble begins when the standard becomes: “if AI touched the process at all, the result is morally contaminated.” At that point, the distinctions disappear.

Generating an entire article, correcting punctuation, brainstorming headlines, translating a paragraph, removing an object from a photograph, making a reference image, or asking whether “whom” is correct in a sentence all get absorbed into the same category.

AI was involved.

Therefore, it is AI.

Therefore, it is bad.

This is an extraordinarily convenient ethical framework because it requires almost no analysis. You do not have to consider intent, scale, harm, substitution, consent, or context. You only need to identify the forbidden substance.

The problem is that technology does not cooperate very well with purity. Machine-learning systems already sit inside search engines, cameras, recommendation systems, spam filters, translation tools, accessibility software, autocomplete, photo processing, and ordinary consumer devices. That does not make someone who intentionally avoids generative image tools but owns a smartphone a hypocrite. It does demonstrate that degree and intention matter, and once degree and intention matter, the binary starts falling apart.

Literally Nobody Asked

This is where the ethical argument becomes a social one.

You are allowed to think generative AI is unethical. You are allowed to refuse to use it, avoid companies that rely on it, argue for regulation, support artists, criticize training practices, reject AI-assisted commissions, and make whatever purchasing decisions follow from your principles. You are even allowed to quietly judge somebody else. A private moral judgment remains one of the few luxuries civilization has not yet managed to monetize. What you are not automatically owed is universal compliance with your ethical framework, and encountering something you dislike does not create an obligation to announce your opposition to it.

A person posts an image. Someone arrives to announce that they do not support AI.

A writer mentions using ChatGPT to organize notes. Someone informs them that “real writers” do not need it.

Someone shares a stupid generated dog. Another person appears to explain global water consumption.

The objection may be sincere, but that does not make the intervention necessary. There is a strange assumption in modern online culture that noticing something morally questionable creates a duty to publicly perform opposition to it. The result is that personal ethics become social enforcement projects, and every mundane interaction becomes an opportunity to demonstrate moral cleanliness.

Sometimes, an AI-generated dog is just an AI-generated dog. You can have principles without turning every principle into a purity test. You can criticize a technology without pretending every use is identical. You can object to specific harms without treating the word “AI” as a complete ethical argument. You can believe somebody else is making the wrong choice without appointing yourself the internet’s compliance officer. The vegans, at least, eventually learned to let people finish dinner. The anti-AI crowd still wants to inspect the menu.