Not Safe for Whom?

Image-generation models block images classified as sexual or violent by labeling them “NSFW” (“Not Safe for Work”). This “safety” is sustained not only by automated classification through code, but also by the invisible labor of content moderators who continuously examine vast numbers of images and decide what is or is not harmful. In this work, I used Python to intervene in the generation process as images evoking queer bodies and forms of intimacy were being excluded, halting the process just before their complete erasure. I then layered painting medium and *gofun*—a traditional Japanese shell-white pigment—onto these fragments and burnished their surfaces with agate, breathing life back into images that were on the verge of disappearing. The labor of making someone invisible, and the labor of depicting again those who have been made invisible. In the age of AI, who is included in the promise of “safety,” and who is excluded?