Texas professor fails half his class after ChatGPT falsely claims their essays

If you want to catch a cheater, asking the AI if it did it is a terrible plan. But one college professor did exactly that, with disastrous results.

A group of seniors at Texas A&M University-Commerce who had already walked the stage at graduation found their diplomas suddenly placed on hold. Their agricultural sciences professor, Dr. Jared Mumm, flunked more than half the class after deciding their final papers were generated by artificial intelligence.

His high-tech detection method? Copying and pasting the students' essays into ChatGPT and asking the bot if it wrote them. The bot happily claimed it did.

Even the professor's dissertation failed the test

The story broke when a Reddit user named DearKick posted an email Mumm sent to the class. In the note, Mumm misspelled the software as "Chat GTP" and explained that he had run everyone's last three essay assignments through the bot twice. If the chatbot claimed the text both times, the student received a zero.

Panicked students who had never used AI tried to defend themselves. Several sent Mumm Google Docs with edit histories and timestamps proving they wrote the essays themselves. According to DearKick, Mumm ignored them at first, leaving remarks in the grading portal: "I don't grade AI bullshit."

Reddit users quickly turned the tables on the professor. One user found an excerpt from Mumm's doctoral dissertation on pig farming and fed it into ChatGPT with the same prompt. The bot responded that the text indeed contained characteristics consistent with AI generation. When users submitted Mumm's angry accusation email to the bot, ChatGPT took credit for that too: "Yes, I wrote the content you've shared."

Chatbots don't keep receipts

The fundamental problem here is that ChatGPT was never built to detect AI-written text, let alone remember its past outputs. Large language models predict likely text; with minimal nudging, ChatGPT will readily claim it wrote classic literature like Crime and Punishment.

While dedicated AI detection tools exist—such as Winston AI or Content at Scale—even OpenAI's own detection tools have struggled with accuracy. Using a conversational chatbot as a lie detector simply does not work.

As the story went viral, Texas A&M-Commerce stepped in to contain the fallout. The university issued a statement clarifying that no students had officially failed or been barred from graduating, and that officials were investigating while drafting formal AI policies for the classroom.

According to updates from DearKick, at least one student was exonerated and received an apology after showing time-stamped files. However, two other students admitted to using ChatGPT earlier in the semester, muddying the waters for everyone else.

Why it matters

This incident is hilarious on the surface, but it points to a very dark reality for anyone producing original work in the AI era.

For students and knowledge workers, version history is now your defense attorney. Writing inside tools with verifiable edit trails, like Google Docs with revision histories turned on, is no longer optional. When managers or teachers lack tech literacy, you may be forced to prove a negative.

We think this mess highlights the sheer panic happening in higher education. Instead of creating clear guidelines on what constitutes acceptable AI assistance, institutions have left individual instructors to police the tech themselves. Mumm's mistake was treating a probabilistic text generator as an infallible source of truth.

Until schools and workplaces roll out clear, evidence-based AI policies, we expect more false positives, damaged reputations, and absurd witch hunts led by people who do not understand the software they are trying to regulate.