Stanford used AI to erase a student from a campus banner photo

A Stanford student found out that he had been edited out of a photo he was in. The university’s housing office had run the picture through a generative AI tool, and on the banner he was someone else.

What was changed

The original photo showed three students at a 2024 Lunar New Year dinner, and it had already appeared in Stanford materials without edits. This fall, Stanford’s Residential & Dining Enterprises used generative AI to rework it for a welcome banner in a student housing center.

According to The Stanford Daily, the edit turned the students’ clothes into Stanford merchandise, slimmed two of their faces and “entirely changed one student’s race, gender and appearance.” That student, senior Billy Ramirez, was replaced with an AI-generated Black woman, NBC News and the SF Standard report.

“Seeing my identity changed and being left out of the picture made me feel, in a way, silenced and erased from a representation that was supposed to include me,” Ramirez told The Stanford Review.

Stanford’s response

Stanford says its policy prohibits using AI to produce or alter images of Stanford people, and that “the alteration and the lack of disclosure in this case violated that policy.” The banners came down. A Stanford statement reported by NBC News said “this was wrong and should never have happened.”

Why a rule was not enough

The notable part is that the rule already existed. Stanford had a written prohibition on exactly this kind of edit, and the image still went through production, printing and display anyway. The check that caught it was a friend sending the student the edited photo.

Generative editing has made it trivial to change who appears in a real photograph, and the result carries no visible seam. In a promotional banner, where nobody expects to verify faces, a swap can stay unnoticed until someone who knows the people sees it. The open question for institutions is whether written AI policies can work without a disclosure and review step that sits before publication rather than after a complaint.