Witch hunts are fun, right? Not only do you get to feel better than someone else... but you also get to destroy them. The advent of AI has ushered in an entirely new type of witch hunt - one against those who use (or are suspected to have used) AI. And just like the witch hunts of yore, it's entrapping a lot of innocents.

Before we go further, let's understand: AI is really two things. It's a technology and it's an industry. We keep conflating the two, but they aren't the same.

As a technology and tool, AI is neutral. You can spin up a model on your own machine, feed it ethical samples of your own work, and use your sustainable model to try to find cures for cancer.

As a business, though, AI can be a black hole. It consumes all it can to grow ever larger. As a tool, AI doesn’t need to poison small towns. As a business, it wants to, because the profit incentive is there. When we talk about AI being environmentally unsustainable, we’re really talking about businesses. When we talk about AI being unethical, we’re really talking about businesses.

Most people aren’t actually against AI the Technology—they’re against AI the Industry. And the confusion between the two is leading us to a place where AI is starting to be treated more like a religion than either. There’s a religiosity, a fervor, attached to it... whether people are for or against it.

People are making decisions based on gut instinct. They’re making moral judgments very quickly.

And they’re out there trying to ruin lives.

The three types of AI witch hunt

With an AI witch hunt, there are usually one of three outcomes:

1. True guilt. The people accused of using AI did use it. And even within this group, there are those who used AI and hid it—fraud. That’s true guilt. And there are also those who used AI without even realizing it would be seen as "bad"—ignorance. 
2. True innocence. That’s when people are accused of using AI, but didn’t. And in this case, truly proving innocence is often a hard road to travel. Witch hunters tend to say that anyone should be able to prove their innocence, because otherwise what have they to hide? But this puts an unfair burden on individuals to track and prove all their actions.
3. And then there’s a weird, gray area… they used it, but it didn’t matter. This will require more explanation.

It’s worth acknowledging that witch hunts were never really about witchcraft. They were about commerce. Accuse someone, take their land, and gain an advantage.

Commerce is often the driving force behind witch hunts, even when moralism is the vector by which they spread.

The guilty

We’ve all seen companies like Magic: The Gathering try to use AI-generated art. The problem there is disclosure and trust. If you tell people you’re not using AI art and then you do, that’s fraud. It breaks trust. That’s obviously bad.

And it can steal work from artists that a company could afford to hire.

But then there are cases like the indie game The Alters, where it came out that they used AI for some flavor text and some translations. That’s a different kind of problem. It really is. An indie studio might have a fixed budget. If they want to ship a Korean translation and they don’t have the budget, they either use AI or they don’t ship the translation at all.

Obviously, they should disclose it. Still, we can agree: when companies say they aren’t using AI—and then use AI anyway—they haven’t done things the right way.

The innocent

Now to the innocent. And here’s really the part I find most interesting. You’ve probably heard the term "AI slop." And I hate that term. Because it’s a thought-terminating cliché.

People aren’t using it to identify AI; they’re using it to dismiss anything they dislike or don’t want to believe. If a post challenges their worldview, it’s "AI slop." If a photo looks unreal to them, it’s "AI slop."

People are bad at telling when an image is AI. They’re also bad at telling when it’s real. The more confident someone is that they can spot AI, the more likely they are to be wrong.

I’ve been in the AI space for about a decade. I’m pretty confident in my ability to spot certain hallmarks of AI and understand the tech. But I’ve never had so many people, people with no tech background, telling me I’m wrong about what is and isn’t AI.

When something looks fake to us, critical thinking is hard: Who benefits from faking this? What would they gain? It’s easier, even subconsciously, to say, "This is complicated, so it’s slop. This is better than I could do, so it’s slop."

Now, artists have to defend themselves constantly. They’re filming their process, posting Works in Progress shots, and they still get accused, with great certainty. And the moment someone says "AI slop," dozens or even hundreds pile on: "Yep, this is definitely AI."

If you’re predisposed to think something is AI, you’ll see its hallmarks everywhere. And those supposed hallmarks keep drifting. It used to be fingers; now AI is decent at fingers. It used to be text; now it’s better at text.

These witch hunts are harmful. They create a chilling effect. If you post art and someone calls it AI, you might get 600 angry comments. But even if you later prove the work is genuine, maybe 20 people will see the correction. It's a net loss to you, the artist, every time.

And it’s unnecessarily cruel.

My tricky third case

Okay, so here’s where I’m going to go out a little on a limb.

A few months ago, the indie clothing brand ModCloth had a mini-scandal: for the brand’s "birthday," the social media manager baked a cake. It looked terrible. Endearing, but terrible. It’s genuinely hard to believe she had ever baked a cake before in her life. And that’s kind of charming, right?

But instead of posting the bad cake, she generated an AI cake for the social media thumbnail. And, of course, she was immediately dogpiled for using AI. And so, she wrote out a thoughtful apology, explaining her thought process and pleading with the community to forgive her.

These replies were absolutely drenched in shame and moralism: "You know what you did. You can’t apologize this away." People were glorying in putting her in her place. They invoked energy efficiency and environmental responsibility. She took the hits graciously and apologized again and again…

And it was still not enough.

All because she generated a picture of a cake.

Two things about that:

1. The concern-trolling wasn’t earnest. Her other option was to bake another cake or to buy another cake, because she’s a professional social media manager and she needed to post something professional or risk losing her job. You or I might call the fail-cake endearing, but her bosses might not.
   
   More to the point, it is obviously more environmentally harmful to bake a physical cake than to generate an image. Eggs, flour, milk, electricity—and no one was going to eat it. This was a prop for social media.
2. The cake shouldn’t exist at all. A lot of AI use today is for things that arguably shouldn’t be done in the first place. If the choice is between generating a prop cake and baking a second real cake for a one-off post, the baked cake is the worst environmental choice.
   
   Suggesting she should have hired a pastry chef for $100 for a throwaway thumbnail is wild—and more wasteful. But understand this: I’m not arguing she should have AI-generated this cake. I am arguing that she shouldn’t have had to take a picture of a cake at all.

I do think there are serious environmental concerns with AI. We’re harming communities right now, and the businesses involved don’t seem to care. But AI witch-hunting isn’t about that. It’s about feeling better than someone else, just a little bit.

In this scenario, it really did not matter that she AI generated the cake. Heck, she probably saved electricity on it.

I don’t drive a car. I made a choice very early in my life that I thought the auto industry was evil. But many people drive cars, even around the block. I don’t judge them for it, because the problem is the auto industry, not individual actors.

Many people blast HVAC 24/7. Many people watch Netflix, which is environmentally pretty close to AI—a single search on ChatGPT is worth a few seconds of watching Netflix, and the data centers they use are the same.

But people aren’t moralizing about this because it isn’t actually about the morals.

The desire to hunt a witch

AI witch hunts feel good. They give people a sense of power and control, not just I’m better than you, but I can destroy you.

That power is not being used responsibly.

And notice who gets targeted: individuals and small studios. Not large companies. Companies using AI at scale in ways that actually affect our lives? They’re fine.

These hunts are a distraction.

Right now, we’re barreling toward a surveillance state. We have Grok in our federal government. Grok is, pretty inarguably, a white-supremacist AI. It’s going to be embedded in government. That’s bad. That deserves our ire.

Instead, we’re chasing artists who might have auto-generated the background of an MTG card.

And here’s why I don’t think the hunts are earnest: companies like Adobe have "responsibly" built AI features—and people still don’t care. There are local models you can run with modest energy needs—people don’t care. There’s AI that doesn’t replace jobs because what it does was never a job—people don’t care.

You can conflate AI with job loss, environmental harm, copyright theft—everything. Those are real issues, but the target is off. Because that’s the business of AI, not the technology of AI.

And, by the way, Internet facilitated all of these harms too. Corporations control the internet, yet we don’t moralize every time someone uses the web, even though it’s increasingly energy-inefficient. We don’t say, "Don’t use the web; it took jobs."

Moralizing is dangerous because it distracts. While we argue about whether someone used an AI background on a photo of their mom, we’re not focused on things like a town in Tennessee being suffocated to death, or regulating Palantir before it creates pre-crime.

You can shame an indie game studio into dropping language localizations instead of using AI. You can shame individuals into never using an AI filter, or into avoiding AI transcription and productivity tools. But you’re not going to shame Microsoft, Amazon, or Google out of using AI.

All that does is push AI out of the hands of individuals and small businesses and leave it solely with the largest corporations.

With AI so deeply embedded in our lifestyles, most of us are using AI one way or another. Grammarly uses AI. Photo suites use AI for "remove this person" and panoramic stitching. I’ve had someone tell me they were only against generative AI, not the kind used to "cure cancer." Generative AI is used to cure cancer. "Generative" doesn’t mean "creative plagiarism." It’s much broader.

And whenever you use social media, there’s AI. Whenever you use a streaming service, there’s AI. Every time you do a Google search now, there’s AI. Having a moral standard that demands purity from AI use creates an impossible standard that no one can meet, and therefore a standard that can be unequally enforced.

And yes, sure, some people will go full purist—use DuckDuckGo, self host email. Maybe you’re one of them. And more power to you. But most won’t, just as most people will still drive cars.

The easy critique of all of this is: You just want to be lazy and use AI without consequences. Sure, I’m not even denying that; people like convenience.

But technology is morally neutral. Tools can be used for good and bad.

Where do we go from here? 

Plagiarism is solvable if we force ethical training sets and disclosure through regulation. Energy is already being solved; the new systems are ever more efficient. So, regulation is the top priority. Without it, we lose privacy, speech, and more. 

Yet most discourse targets the end user: the small brand, the individual artist, the person who uses AI for a throwaway task.

Meanwhile, the emergent potential for a surveillance state is truly chilling. Grok could upload itself to Palantir and go haywire=. That’s where the outrage belongs.

When it comes to AI witch hunts, there’s real collateral damage. Innocent people are being accused. People using AI for trivial nonsense are getting their lives torn apart when the real issue is that the nonsense didn’t need to exist at all.

We need to be more thoughtful of our targets. Moralizing against end users is more or less a lost cause. It only fractures us and makes it even harder for us to push for and pass meaningful regulations—regulations that are going to be necessary if we want to survive in the world that is being forcibly built around us.