When ChatGPT Enters the Exam Hall: AI-Assisted Fraud in India's Professional Education

A Hidden Phone, a Washroom, and ChatGPT

During a biochemistry mid-semester examination at All India Institute of Medical Sciences (AIIMS), New Delhi, faculty members noticed an unusual pattern: an abnormally large number of students were requesting restroom breaks. When staff investigated, they discovered a mobile phone hidden inside the washroom. Approximately 50 to 60 MBBS students had allegedly been queuing up to photograph questions and retrieve answers from ChatGPT before returning to their answer sheets.

AIIMS Delhi cancelled the examination immediately and scheduled a re-examination. The incident, widely reported in May 2026, sent a signal through India's medical education community: AI-assisted cheating had moved from school corridors into one of the country's most prestigious medical institutions.

This was not an isolated event.

The Gadchiroli NEET Cases

On May 3, 2026, the same day that 22 lakh NEET UG candidates sat for the national medical entrance examination across 556 cities, two candidates at examination centres in Gadchiroli, Maharashtra, were found with mobile phones and evidence of AI tool usage during the exam.

Centre heads filed complaints at Gadchiroli police station. Cases were registered under Section 3(viii) of the Public Examinations (Prevention of Unfair Means) Act, 2024 — a law enacted specifically to address organised cheating in national examinations. Phones and digital evidence were seized for forensic examination, and an Assistant Superintendent of Police was assigned to investigate whether a larger network was involved.

Combined with the AIIMS incident, the Gadchiroli cases illustrate a structural shift in how exam fraud operates in India.

What Changed: From Paper Chits to Live AI Queries

Traditional cheating methods — handwritten notes on palms, paper chits, proxy candidates — are visible, physical, and slow. Modern AI-assisted cheating is fundamentally different in three ways.

Real-time generation. A student who types a question into ChatGPT in a washroom receives a coherent, contextually plausible answer within seconds. Unlike a copied note, the AI-generated answer is tailored to the exact phrasing of the question, making it harder to flag as generic or pre-written.

No prior preparation required. Paper-leak cheating requires obtaining the question paper before exam day, which requires an insider. AI-based cheating requires only a phone and a working internet connection — both increasingly small to conceal, and increasingly difficult to detect once inside a centre.

Scale and organisation. The AIIMS incident involved an estimated 50 to 60 students coordinating a single hidden device in a washroom. The Gadchiroli cases suggest individual candidates acting independently. Both patterns are worrying: the first shows that coordinated exploitation is possible even in heavily supervised professional environments; the second shows that individual opportunism is also rising.

Data from UK universities — which tend to lead India by 12 to 18 months on technology adoption trends — shows AI-assisted cheating cases rose from 1.6 per 1,000 students to 5.1 per 1,000 in a single year. Indian higher education institutions should treat this trajectory as a near-term planning horizon, not a distant warning.

Why Professional Education Is Especially Vulnerable

MBBS and other professional entrance examinations carry unusually high stakes. A few marks determine admission to premium medical colleges, career trajectories, and lifetime earnings. This intensity creates powerful incentives for candidates to take risks they might not take in lower-stakes contexts.

At the same time, the examination format most common in professional education — long paper-based examinations in large halls — concentrates risk. When hundreds of candidates share a venue, and invigilation relies primarily on human supervision, the ratio of invigilators to students rarely allows for the close observation that AI-based cheating demands.

The AIIMS case is particularly instructive because it occurred inside a controlled campus facility, not an outsourced examination venue. If AI-assisted cheating can penetrate a premier national institution's own examination infrastructure, the challenge for universities using third-party evaluation centres is even more acute.

The Surveillance Response: AI vs AI

India is not standing still. Bihar's Police Subordinate Services Commission deployed a comprehensive AI surveillance system for its Sub-Inspector examination, covering 6.60 lakh candidates across 613 centres in all 38 districts. The system used facial recognition and eye-movement tracking connected to a central command room through 16,500 CCTV cameras.

The Bihar deployment achieved real deterrence — impersonation fraud, which had historically inflated SI examination malpractice, was significantly curtailed. Facial recognition at entry prevented candidates from substituting ghost writers; eye-tracking inside halls flagged suspicious gaze patterns for human review.

But AI surveillance addresses cheating within the examination hall. It does not solve the washroom problem, or the growing sophistication of concealed devices — increasingly disguised as glasses frames, watch components, or hearing aid hardware.

A comprehensive answer requires rethinking not just surveillance, but the examination format itself.

What Institutions Can Do

Move evaluation away from the scene of cheating. When answer sheets remain at examination venues or are distributed to evaluators at home, they are exposed to tampering at multiple points. Digital evaluation systems that scan scripts immediately after collection and upload them to a secured central platform break the physical chain. A script that never leaves the scanning room cannot be altered between writing and evaluation.

Design AI-resistant assessments. Examination bodies are increasingly recognising that any question with a single correct factual answer can be answered by a language model. Questions that require candidates to reason from partial information, identify inconsistencies in a presented scenario, or synthesise data from multiple sources are harder to outsource to AI. The shift to competency-based assessment under NEP 2020 naturally pushes evaluation in this direction.

Create audit trails at every stage. From the moment a candidate submits a script to the moment a mark is uploaded, every action should be timestamped and logged. Digital evaluation platforms that record evaluator interactions — which script was opened, when, for how long, what mark was awarded — create a forensic record that both deters manipulation and supports post-incident investigation.

Enforce the Public Examinations Act. The 2024 Act criminalises organised cheating with penalties of up to ten years' imprisonment and fines up to one crore rupees for organisers. The Gadchiroli cases are among the first applications of this statute to AI-based fraud. Consistent enforcement will establish deterrent precedent.

The Integrity Dividend

The academic community often frames examination integrity as a compliance problem — a burden to be managed. The more useful frame is institutional credibility. A degree from an institution with a documented history of examination fraud is worth less in the job market, in postgraduate admissions, and in NAAC assessment, regardless of the individual student's actual competence.

Institutions that invest in robust examination infrastructure — AI surveillance, digital evaluation, competency-based question design, tamper-proof audit logs — are not just preventing cheating. They are protecting the value of every certificate they issue.

The AIIMS and Gadchiroli cases are early data points in a trend that will intensify as AI tools become smaller, faster, and more accessible. The institutions that establish rigorous systems now will be better positioned to maintain credibility as the landscape evolves.

---

Related Reading