Veer Narmad South Gujarat University in Surat caught two students using ChatGPT and Google Gemini through smartwatches during an exam, and fined them 10,000 rupees each, double the standard penalty. Months later it caught six more students, two of them MBBS candidates, using ChatGPT through phones and Bluetooth earpieces, same fine, exams cancelled. Then in March this year the university caught close to 400 students cheating in a single exam cycle and, this time, chose not to punish most of them. Around 226 were called before a malpractice inquiry committee, and 14 were sent for counselling instead of a fine or a ban.
Two responses to the same fact, months apart. The first time, the university tried to out-punish the tool: raise the fine, ban the smartwatch, move on. The second time it seemed to admit that would not hold, and asked why students were reaching for AI in the first place. That is a more honest question, but it still stops one step short. Nobody at the institutional level is asking whether the exam itself is the problem. If a smartwatch running Gemini can pass your test, the test is measuring recall of things a phone already knows, not understanding.
I build study products for the same students sitting these exams. InfiNotes and the study companion I have not launched yet exist because the line between AI helping you learn and AI helping you cheat is not the technology. It is entirely about when a student reaches for it: the week before the exam, or the minute during it. A fine changes nothing about that timing decision. A better assessment might.
The honest caveat: I have not solved this either. Nobody building AI study tools has a real answer for how an exam should be redesigned once the thing it tests for can be answered by a watch. Counselling 400 students instead of fining them is more compassionate. It is not a fix.