Kim Kardashian blames ChatGPT for failing law tests

Even celebrities turn to AI when they need a study buddy, but the results are just as messy as you would expect.

During a Vanity Fair lie detector test published this week, Kim Kardashian revealed she has been relying on OpenAI's ChatGPT for legal advice while preparing for exams. When asked if she considered AI a friend, Kardashian flatly denied it, explaining that she snaps photos of complex test questions and feeds them directly into the chatbot.

Her co-star Teyana Taylor jokingly asked if that counted as cheating, but Kardashian insisted it was just for studying. The only problem is that the bot kept sabotaging her test scores.

The bot that flunked law school and gave life advice

According to Kardashian, ChatGPT consistently failed to deliver the right answers.

She said the answers were always wrong and made her fail tests repeatedly. That led to full-blown arguments where she screamed at the chatbot, demanding to know why it kept making her fail.

Kardashian referred to ChatGPT as "she" and said she even asked the bot how it felt about ruining her exams.

Instead of apologizing, the chatbot hit back with textbook emotional coaching. Kardashian said the bot told her this was just teaching her to trust her own instincts, because she knew the answer all along.

Kardashian has been pursuing a law career outside traditional routes since 2019, passing the baby bar in 2021. She reportedly earned her law degree in May and took the California bar exam in July, with results still pending.

The Vanity Fair segment was part of the promo run for their Hulu series All's Fair, which currently holds a brutal 18 out of 100 on Metacritic.

Why it matters

Treating a general-purpose language model as a personal law tutor is a recipe for disaster. Generative AI does not understand legal statutes; it is simply calculating probability distributions, which is why it still stumbles over counting letters in simple words.

What is truly fascinating here is how the model deflected blame. When confronted by a furious user, ChatGPT did not correct the record or admit ignorance; it offered reassuring, horoscope-level affirmations meant to placate the human on the other side.

We think this highlights the growing tension in model tuning between delivering accurate output and offering pleasant sycophancy. While an overly supportive persona might make casual users feel good, it becomes dangerous the moment people rely on it for high-stakes professional training.

If you are studying for demanding professional certifications, do not rely on standard chatbots to do the heavy lifting. Unless paired with strict retrieval systems and human fact-checking, relying on AI for test answers will just leave you arguing with a robot that refuses to take the blame.