Human Modified

Art of FACELESS on labels, looks, and where the studio goes next

Ten days ago, Pinterest put two words on our work: "AI modified." Since then, our weekly reach has gone from around 40,000 to 15,000.

We're not appealing, and we're not upset with the people who scrolled past. If a platform tells you something is fake, it's perfectly reasonable to believe it. Most of us would.

So this isn't a complaint. It's a report from a small studio that's been making things since 2012: what happened, what we think it means, and what we're doing about it. Spoiler: we're not closing. We're changing shape.

What happened

Earlier this year we took our whole Pinterest catalogue down. Our handmade 3D work, built in Character Creator 3 and 5, processed as photography and credited as exactly that, had been labelled "AI modified." We came back recently for one practical reason: the shop.

Then it happened again. The cover of The Hollow Circuit® is two years old. We regraded it in our official brand palette, and that was the only change. The monochrome versions of the same artwork pass without a mark. Add teal and coral, and suddenly it's suspect.

Then a photograph. It's a portrait our director shot about twenty years ago and reworked in Mosh-Pro. Mosh-Pro isn't generative AI. Every part of that frame was shot by a person and processed by a person, using the kind of post-production any working photographer was doing in Photoshop in 2006. The original has never been online, and it's going to stay that way.

Two small words

It's worth reading the label slowly, because it's been written very carefully.

It doesn't say "flagged by an automated classifier." It doesn't say "may contain," or give a confidence level, or name a tool. It doesn't say how much was changed, or whether a person ever looked. It says "AI modified," and stops.

Most people will read that as a verdict, and fair enough. In everyday English, it sounds like a finding: someone checked, and this isn't quite what it claims to be.

Pinterest hasn't said whether labelled work is shown to fewer people, and to be fair, it doesn't need to. It offers users a setting to see less "AI" in categories like art. The label quietly does the rest.

Meanwhile, the platform is full of extraordinary images carrying that badge, whatever their actual history. Plenty of genuinely generated filler carries no label at all. Once the badge is on, though, the quality of the work stops being the conversation. The label becomes the conversation.

Slop, and other words

We've said for a while that "slop" is a slur, and we'd like to explain why, because it isn't hyperbole.

Slurs tend to work the same way. They take one trait, real or assumed, and make it the whole person. That's how slag and slut work, and it's how flid and spastic work. "Slop" does it to a piece of work and, by extension, to whoever made it. "Spastic" began as a clinical term before it became an insult, which is part of why the Spastics Society became Scope.

"AI modified" is the polite, institutional relative of "slop." It's clinical and neutral-sounding, applied by a system rather than shouted in the comments. It sounds like a fact, which is exactly why it does more damage than an insult.

We're not saying the two hurt the same. We're saying they're built the same.

And if "AI modified" covers grading, retouching and stylised post-production, then it covers most of what professionals have done since the darkroom. Almost every image you've ever admired was modified by somebody.

We're not anti-AI. We never have been.

This is the part that usually gets lost, so we'll be plain. Art of FACELESS has never said AI is off limits. We run our own local models, on our own machines, in our own studio. They don't harvest anyone's work, and they don't send anything to the cloud.

That's why the label leaves us in such an odd position. Either way, the artist loses.

If the label is wrong, it's a public mark on someone's work that shouldn't be there. If it's right, because someone used a tool to make work they couldn't otherwise make, then the label is singling people out for the tools they need. Either way, the badge doesn't help anyone understand the work.

From our director, Lloyd Lewis:

I'm disabled. I live with Secondary Progressive MS. When our handmade 3D work was labelled "AI modified" earlier this year, it landed harder than I expected, because I recognised the demand. Disabled people are asked to prove they're real all the time: real enough to need help, but not so real it makes anyone uncomfortable. Here was the same question, asked of my work.

Shaming someone for using AI to make art is a lot like shaming a disabled person for wearing smart glasses that describe the room to them. The tool isn't the problem. I won't be justifying my wheelchair, whatever it's made of.

Nobody speaks for us. Including us.

Spend a little time in the comment threads, and you'll find a lot of non-disabled people speaking on behalf of disabled people. They explain what disabled artists need and don't need, which tools are acceptable for them and which aren't. They mean well, mostly. But it's still a conversation about us that we weren't invited to.

Nothing about us without us. That principle is old, and it's good, and it applies here too.

It also applies to us. We don't represent the disabled community. We write as disabled artists, for ourselves and nobody else. Other disabled artists may see all of this very differently, and they have every right to. We aren't the voice of disabled artists, and we'd be uneasy with anyone who claimed to be.

We're Art of FACELESS. That's all.

It's about people and money

None of this is really about AI. It's about how people behave in a crowd.

Much of what lands on artists right now is a purity pile-on. A lot of it is groupthink, and some of it isn't even people: it's bots, farming outrage at whatever moves. It doesn't much matter which, because the damage still lands on real work, real names and real income.

Underneath that is something quieter: money. Making art takes materials, software, tuition, and hours of practice that most working people don't have. It takes permission, too: growing up somewhere wanting to be an artist wasn't treated as a joke. Even the time and energy to spend an afternoon online arguing about AI art is a luxury.

Plenty of working-class people don't have it. Plenty of disabled people don't have it. Plenty of older people don't, or people who left school at sixteen, or people who never went to university. Many of them have had a story, a song or a picture in their heads for decades.

If you were lucky enough to always be able to draw, that's a wonderful thing. We'd just gently suggest it's a gift, not a credential, and a gift is a strange thing to use as a weapon against someone who never got one.

The look has been taken

Here's something we don't hear said often enough. Generative models didn't just flood the market. They absorbed a visual language.

Grain, glitch, smooth skin, glossy highlights, neon complementaries, teal against coral: artists spent decades developing those looks. Now machines can produce them in seconds, and people have understandably started to assume that anything with that look came from a machine. Classifiers do it with pixels. People do it with their eyes.

If your style is one the models learned well, you're suspected by resemblance. That's a genuinely new problem for working professionals, and nobody has a good answer to it yet, including us.

A question about process videos

Artists are increasingly told that the best defence is to film themselves working. The time-lapse, the layer stack, the sketch, the hands on the tablet. Show your working.

We'd like to ask, gently, whether that might backfire.

Around 2012 and 2013 we stopped posting original work and stock on big platforms, because we'd watched what happened to everything that was posted. It was copied, collected, and eventually became part of the training data for the models that now imitate it.

Process footage goes further. It hands over the method as well as the result: the decisions, the order of operations, the craft that took years to learn. That's about as useful as training data gets. It also adds to the storage, the bandwidth and the energy the internet already burns.

We're not against AI, as we've said. We're just not keen on our work being taken without asking. So we won't be filming. No face. No performance. No proof required.

Where we've landed

A few honest positions, offered with as much good humour as we can manage.

To artists posting original work to prove a point: we understand why. We'd only say that the label doesn't check how good you are before it lands. It might be worth asking what you're handing over, and to whom.

To anyone running art through a detector before deciding whether they like it: we'd invite you to just look at it. You're a much better judge than the software, and a lot more fun.

To the platforms: if you're going to publish a verdict on someone's work, it seems fair to show the evidence and the error rate. A form isn't an answer.

To the EU: we understand why labelling duties exist, and we share the concern behind them. But writing them into law leaves platforms reaching for the cheapest tool that technically complies, and right now that's probabilistic guesswork with a badge on it. Provenance isn't the same as pixel statistics. We'd love the law to know the difference.

The bigger risk

If original work gets marked as fake the moment it goes online, professionals will sensibly stop putting it online. Then what's left will be mostly generative. It would be a strange ending: the most passionate campaign against AI art helping to build an internet where it's almost the only art left.

We don't think that has to happen, but it's worth saying out loud while there's still time to change course.

What we're doing

Art of FACELESS is founded and run by a disabled director with Secondary Progressive MS, who has worked professionally since the late 1990s. The studio has been around since 2012, and in nearly fifteen years we've rebuilt how we show up more than once. This is another of those times.

We're not closing. We're adapting.

We're staying on Pinterest, as artists. The "AI modified" label is now the subject of an art project, and Pinterest is where it will be shown. Every Pin will carry our own disclaimer explaining what the label can and can't tell you, and how it can misrepresent the work it sits on. And yes, every Pin will link to the shop. We're not going to pretend otherwise. Call it artistic licence.

We broadcast; we don't live there. Like every platform we use, Pinterest becomes broadcast-only. We post, and the conversation happens on our own channels.

Originals stay offline. What we share online from now on will be previews, details, lower-resolution versions and abstractions, never the master files. The originals live in print, in our books, and in our own locally run systems.

The shop is open. Nothing changes for anyone who buys from us.

We're becoming a studio in a fuller sense. Not only a publisher that releases things, but a place where things get made, with people. Radical in the old meaning of the word: going back to the root. That means opening the door to working-class people, disabled people and older people who've always had the ideas but never had the means, the time or the permission. No portfolio required. No proof required. Nothing about us without us.

We'll share more about that over the coming weeks.

Because in the end, the question was never really "Is it AI?" It's this: who gets to imagine?

We'd like the answer to be "everyone."

Art of FACELESS
Punk publishing. Run by disabled people. No proof required. Since 2012.

Notes: Pinterest describes its "AI modified" label in its Gen AI labels announcement. Its "see fewer" controls for AI content in categories including art are covered by TechCrunch. Pinterest has published nothing on how labelled Pins are ranked (Versely). WIRED found AI-generated Pins carrying no label. Our case file on palette-based false positives is The Algorithmic Smear.

The Hollow Circuit®, Hyperstition Architecture®, The Veylon Protocol® and Cognitive Colonisation® are proprietary intellectual property of Art of FACELESS.