Stanford apologizes: it used AI to change a student's race and gender on a poster, despite its own ban

By EdSource · September 25, 2026.

Stanford has acknowledged that a promotional poster displayed on campus this fall was altered with artificial intelligence: it replaced a Latino student with a Black woman, slimmed down and distorted the features of his two companions, and played up the university's logo on their T-shirts. The institution apologized after the Los Angeles Times broke the story.

The original photograph, taken in 2024 by Stanford's housing and dining department, showed three students holding plates of noodles during a Lunar New Year celebration. One of them, Billy Ramirez, told the student newspaper The Stanford Review that he was stunned to see himself replaced by someone else on the poster the university put up around campus.

Stanford spokesperson Charlene Gage confirmed to the Los Angeles Times that AI was used to alter the image "in violation of university policy." As she explained, that policy explicitly prohibits using AI to produce or alter images of Stanford people, events, research, facilities or achievements, and both the alteration and the failure to disclose it broke the rule. Gage added that the university is already working on providing additional staff training and on tightening the review of materials before they are published.

What is striking about the case is not that Stanford lacked rules: it had them, and they were explicit. What is striking is that someone used a generative editing tool to "improve" a marketing photo, replaced a real person with another of a different race and gender, and no one stopped it before the poster was printed and hung on campus. The source does not say who made the alteration.

That is the underlying problem, and it is not unique to Stanford. Generative AI retouching features (Adobe's "generative fill," similar tools in Canva and in a dozen marketing apps) have been built into corporate communications workflows by default for more than two years now. Asking an AI to "slim down" someone or swap out a face takes seconds; reviewing it with human judgment before printing takes longer, and someone has to take on that job. When a policy lives only in a document and not in the actual approval workflow, it protects no one.

As industry context, US universities have racked up more than two decades of scandals over manually retouching the racial diversity of their brochures and web pages. What changes with generative AI is not the temptation --that has always been there-- but how easy it is to give in to it without leaving any trace of a conscious editorial decision: swapping a face no longer takes an expert Photoshop editor, it takes a prompt and hitting "generate."

Our read is that this is exactly the kind of short-term friction to be expected when generative AI slips into an organization without the governance that should come with it. It is not a failure of the model --no AI "decided" to commit a discriminatory act--, it is a failure of process: no one signed off on the change, no one reviewed the output, no one asked why the original photo was no longer good enough. It is the same pattern we keep seeing at companies that deploy AI agents without first redesigning who decides and who audits what the machine produces: the technology is almost never the real bottleneck, the operating model around it is.

In the long run, generating and editing images with AI will be a normal part of how any institution --large or small-- produces its materials, drastically lowering the cost of what once required a full design studio. That cost collapse is, at bottom, another form of the abundance we champion at Zendoric: more creative capacity within everyone's reach. But the Stanford case is an honest reminder that this abundance only pays off if it comes, from day one, with rules on disclosure, traceability and human review that are genuinely enforced, not merely written down in a policy manual.