Gallery Entry
Not Safe For Workflow
By Ella Josephine Esque · AT
“Not Safe For Workflow” is a self-portrait that claims to be a prompt. Ella Esque poses as AI-generated and holds up the prompt that supposedly produced her. Except the prompt is made of cardboard, and she is holding it in her own hands. While everyone else is busy proving they didn’t use AI, she claims the opposite. The card opens like any routine specification, “female post-AI artist, realistic skin texture”, and then slips: “sudden defiance --v 8.2 --resistance expected”. Even rebellion has become a parameter, something you can order and expect to be delivered. Directly below, the system answers in red: “AI moderator believes this artist violates our community standards.” It flags the only element in the picture that cannot be re-rolled. The title borrows from NSFW, the label content moderation exists to catch. Nothing here endangers anyone’s morals. What the image endangers is the workflow: the smooth delivery of a person exactly as specified. Scraped into a training set, it turns into a deliberately mislabeled data point, a photograph captioned as machine output. Data poisoning as digital self-defense, with a polite smile. And it is not an innocent photograph either. Photoshop’s AI tools extended the background, and the file’s Content Credentials now flag the portrait as containing AI-generated content. The work does not pretend to stand outside the system it mocks. It sits exactly on the line that labels, detectors and moderators are trying to draw, and smudges it. “Not Safe For Workflow” was created in the context of Ella Esque’s text and upcoming keynote “Destroy What Calculates You” (“Mach kaputt, was dich berechnet”) at USTP St. Pölten in November 2026. The German title riffs on Ton Steine Scherben’s 1970 anthem “Macht kaputt, was euch kaputt macht” (Destroy What Destroys You). The keynote is about the move from acceptance to sabotage: methodical, subversive ways of working with AI that throw sand into the gears instead of oiling them. It asks how artists and designers can protect their data, and their unpredictability, from systems built to calculate them.