Who Made This? AI Assumptions Are Changing How We See Art

Two things happened this month that, taken together, say something important about where AI and creativity actually stand right now — not in the hype cycle, but in the messy, human reality of making and viewing art.

When Digital Work Gets Mistaken for "AI Slop"

On August 6, 2026, the Walker Art Center in Minneapolis opened Hydricosmic Litanies, the first major museum solo exhibition for Brooklyn-based artist, architect, and designer Olalekan Jeyifous — a trained architect and Venice Architecture Biennale Silver Lion winner who imagined a speculative society in Minneapolis that evolved around the ecology of the Mississippi River. The work included digital collages of the cityscape and a series of 3D-printed sculptures depicting the industrial and cultural artifacts of this imagined riverine world.

Nine days after it opened, a Reddit user posted an "AI slop alert," saying they were "disappointed to see so much AI garbage" in the show and claiming that "nowhere in the exhibit did they let guests know that Jeyifous used AI." The accusation spread. The Walker Art Center was soon adding context to the exhibition after visitors mistook the digitally rendered artwork for generative AI.

Here's the thing: Jeyifous did not use generative AI to create the work. His highly detailed digital illustrations — built through years of craft in architectural rendering and 3D design software — looked unfamiliar enough to viewers that their brains filed them under the only category that seemed to fit: AI-generated.

The episode is a mirror held up to all of us right now. We've trained ourselves — understandably — to be skeptical of images that feel too detailed, too synthetic, too polished. That skepticism is healthy in many contexts. But when it misfires on a decade-long practice by a celebrated artist-architect, it reveals something uncomfortable: the cultural category of "AI slop" has grown so large in our imagination that it's swallowing work it was never meant to describe.

For creators here on Sunporch, this is a real concern. If your AI-assisted work is thoughtful, intentional, and refined, it may still be dismissed before anyone reads your process notes. And if you don't use AI but your aesthetic sits in certain zones — hyper-real, digitally precise, algorithmically clean — you may face the same unfair shorthand.

The Science Behind "Who Made This?"

Almost simultaneously, on August 18, a research team at MIT's Computer Science and Artificial Intelligence Laboratory published findings that complicate the question from the opposite direction — not is this AI? but whose work trained it?

New work from MIT CSAIL suggests that for models trained on large datasets, the question of which artists' work influenced a generated image may often have no answer — not because the tools for finding it are inadequate, but because the connection itself has disappeared.

The researchers identified a phenomenon they're calling attribution decay. The more data a generative model is trained on, the less any individual training example matters to any particular output. At large scales, you can often remove any single image from AI training data, every image by a given artist, or every photograph of a given person, and the generated output won't change appreciably.

Surgically removing examples from a model revealed this dissolving link, meaning tracing the origin of an AI-generated style becomes increasingly difficult — researchers found that larger datasets do not necessarily strengthen the connection between source material and output, but rather obscure it.

This matters enormously for the ongoing legal and ethical debates around AI training. When an AI image generator produces a portrait, the question of whose work went into it sits at the center of lawsuits, licensing deals, and proposed regulations worldwide. The MIT findings don't resolve those debates — they arguably make them harder. If removing every image by a given artist from a training set doesn't change what the model produces, what does "influence" even mean at scale?

Two Incidents, One Bigger Conversation

These two stories might seem unrelated, but they're orbiting the same question: how do we assign authorship, intention, and responsibility in the age of generative AI?

In the Walker case, authorship was falsely attributed to AI when none existed. In the MIT study, authorship was found to be untraceable even when AI was clearly involved. Both outcomes are disorienting — and both demand more nuance than our current cultural shorthand allows.

For AI creators specifically, a few things are worth sitting with:

Your process statement is part of the work. In an environment where audiences are hair-trigger skeptical, context isn't a defensive footnote — it's part of what makes the work legible. Describe your tools, your decisions, your iterations. Not to justify yourself, but because that story is genuinely interesting and relevant.

"AI-assisted" covers a huge spectrum. There's a vast difference between typing a prompt and pressing generate versus spending weeks refining outputs, training custom models, or using AI as one layer in a deeply human-directed process. The Jeyifous incident is a useful reminder that even non-AI work can be misread. Your job as a creator is to close that gap with context, not defensiveness.

The authorship question isn't going away — but it's also not fully settled against creators. Legal developments are beginning to establish that AI video and image outputs qualify for copyright protection when significantly modified by humans — though training data ownership remains actively contested. The MIT findings may actually support creators in some contexts, since they complicate the argument that any single artist's work was clearly copied.

What This Means for Your Practice

Demand is rising for creator-first tools that give artists fine-grained control and sovereignty over artistic direction, allowing them to adjust outputs until the work precisely reflects their authentic vision. That demand is good news. It signals that the market — and the culture — is starting to distinguish between passive prompt generation and genuine creative authorship.

Audiences are craving uniqueness and personal meaning, rejecting work that feels standardized or interchangeable — and AI art focused on personal storytelling is becoming a quickly growing area, aiming to push back against concerns about hollowness in generic AI-produced outputs.

The best response to a confusing cultural moment isn't to shrink from it. It's to make work so clearly yours — in its voice, its obsessions, its specific point of view — that the question of the tool becomes secondary to the question of the person holding it.

The Walker visitor who cried "AI slop" at Olalekan Jeyifous's meticulous speculative architecture was wrong about the facts. But the impulse behind their concern — does this work come from a real human perspective? — is a legitimate one. The best answer any creator can give isn't a disclaimer. It's the work itself.

Who Made This? AI Assumptions Are Changing How We See Art

When Digital Work Gets Mistaken for "AI Slop"

The Science Behind "Who Made This?"

Two Incidents, One Bigger Conversation

What This Means for Your Practice

Sources