I've worked at EPFL for over a decade. I’ve reviewed hundreds of applications, mentored students, and fought for fairness in admissions. This year, for the first time, the entire graduate admissions process was handed over to an AI model.
The official line? “It improves efficiency and reduces bias.” The truth? It’s a black box that rubber-stamps whatever the developers trained it to favor.
I raised concerns early. I noticed the model was disproportionately rejecting students from certain regions, undervaluing interdisciplinary backgrounds, and favoring candidates from just a handful of partner institutions. I asked for transparency. I was removed from the review committee.
They told me I was being “disruptive.” My access to admissions data was revoked. I’m no longer invited to workgroup meetings. Just like that—silenced.
Let me be clear: this is not progress. This is academic gatekeeping hidden behind an algorithm no one is allowed to question.
The worst part? Faculty were told to rubber-stamp AI decisions unless there was an “exceptionally clear error.” So much for human judgment.
If you're a prospective student reading this: your rejection may have come from a model that doesn't understand your context, your passion, or your potential.
If they want to shut me up, they’ll have to do more than remove my login credentials. Here's everything I managed to walk away with the day before they kicked me out.