---
title: "India Criminalised Question-Paper Leaks in 2024. Your Hiring Assessment Still Reuses the Same 40 Items."
description: "Proctoring protects the exam session; item security protects the score. What India's 2024 anti-leak law and the NTA reform programme reveal about assessment content risk - plus a seven-step item-security check for high-volume hiring."
canonical: https://www.assessall.com/blog/india-criminalised-question-paper-leaks-in-2024-your-hiring-assessment-still
published: 2026-11-06
updated: 2026-11-06
author: "AssessAll Editorial"
source: AssessAll
---

# India Criminalised Question-Paper Leaks in 2024. Your Hiring Assessment Still Reuses the Same 40 Items.

Proctoring protects the exam session; item security protects the score. What India's 2024 anti-leak law and the NTA reform programme reveal about assessment content risk - plus a seven-step item-security check for high-volume hiring.

Item security is the protection of assessment content — questions, answer keys, scoring rules — from leakage and overexposure. It is separate from proctoring, which protects the administration. A proctored test built on a leaked question bank produces clean-looking sessions and invalid scores, because the candidate no longer needs to cheat in the room.

Indian employers have spent three years buying proctoring. Almost none have bought item security, and the two solve different problems. If your campus drive, BPO walk-in or BFSI frontline screen has run on the same 40 questions since last year, the webcam is watching the wrong thing.

## India already decided this is a serious problem

The Public Examinations (Prevention of Unfair Means) Act, 2024 moved through both houses of Parliament in four sitting days — introduced in the Lok Sabha on 5 February 2024, passed there on 6 February and in the Rajya Sabha on 9 February. What it criminalises is instructive. The first item on its list of unfair means is "unauthorised access or leakage of question paper or answer key" — ahead of helping a candidate during the exam, ahead of tampering with a merit list.

The penalty structure tells you where Parliament thought the risk sat. Individuals face three to five years' imprisonment and a fine of up to ₹10 lakh. A **service provider** — the vendor running the exam — faces a fine of up to ₹1 crore, recovery of the proportionate cost of the examination, and a four-year bar from conducting public examinations; directors or senior managers who connived face three to ten years. Organised crime draws five to ten years and a fine of at least ₹1 crore. All offences are cognisable and non-bailable. ([PRS India summary](https://prsindia.org/billtrack/the-public-examinations-prevention-of-unfair-means-bill-2024))

The Act applies to UPSC, SSC, Railway Recruitment Boards, the National Testing Agency, IBPS and central government recruitment — not to your hiring process. But the reform programme that followed is a free design review. The Ministry of Education constituted a seven-member High-Level Committee of Experts under former ISRO chairman K. Radhakrishnan in June 2024, after UGC-NET was cancelled following a leak on the darknet. Its report, submitted on 21 October 2024, carried **101 recommendations grouped into 46 major action points**, and a high-powered steering committee was set up on 14 November 2024 to monitor implementation. By May 2026, biometric authentication had been rolled out for NEET-UG and more than 99.5% of its centres were government institutions; multi-session examinations and stronger information-security systems were still being phased in, and a further task force under Nandan Nilekani was announced on 26 July 2026. ([Business Standard, 27 July 2026](https://www.business-standard.com/india-news/exam-reforms-k-radhakrishnan-panel-nta-recommendations-paper-leak-126072700556_1.html))

Note the sequence. Identity verification arrived first because it is easy to buy. Content security arrived last because it requires test design, not hardware.

## The evidence: exposure makes items easier before anyone notices

The quiet failure mode is not mass cheating. It is drift.

A study of clinical anatomy examinations at the University of Porto tracked eight 100-item multiple-choice papers taken by roughly 200 students each, in which 4–26% of items per paper were reused from an earlier sitting. Using a deterministic gated item response theory model, the authors estimated the true prevalence of item pre-knowledge at **1.2% to 3.7% of students** — low, and much lower than the 25% who self-reported it.

The more useful finding is what happened to the items themselves. The proportion of candidates answering a reused item correctly rose **3.5% on its first repetition (p = 0.013) and 6.9% on its second (p = 0.036)**. The questions had not changed. They had simply been seen. ([Severo et al., *BMC Medical Education*, 2019](https://bmcmededuc.biomedcentral.com/articles/10.1186/s12909-019-1710-z))

Scale that to a volume funnel. If a handful of items get a few points easier each cycle, your cut score quietly loosens. Pass rates rise, nobody complains, and the people you reject are increasingly the ones who did not find the WhatsApp group.

Adaptive testing concentrates the damage rather than diluting it. As Georgiadou and colleagues note in their review of exposure-control research, "particular items in the item pool may be presented too often and become overexposed, while other items are rarely selected by the CAT algorithm" — and because selection favours items with high discrimination parameters, **the items that leak first are your best ones**. That is why exposure-control methods such as Sympson-Hetter exist: they cap the probability that any item is administered, accepting a small precision cost to keep the pool alive. ([Georgiadou, Triantafillou & Economides, *JTLA* 5(8), 2007](https://ejournals.bc.edu/ojs/index.php/jtla/article/viewFile/1647/1482))

The National Council on Measurement in Education puts the stake plainly: "Test security violations undermine the validity of intended score interpretations and uses, which can lead to faulty decisions based on test scores." Its second recommendation is operational — sponsors and vendors "should analyze examinee response and performance data to identify potential irregularities and possible misconduct." ([NCME position statement, 2019](https://ncme.org/wp-content/uploads/2025/10/NCME_position_statement_on_test_security_03-18-19.pdf))

## A seven-step item-security check you can run this week

This is a design and monitoring exercise, not a procurement one. Work through it in order.

1. **Compute your exposure ratio.** Divide the number of distinct live items by the number of items any one candidate sees. A 40-item test drawing from a 60-item bank has an exposure ratio of 1.5 — every candidate sees two-thirds of everything you own. Write the number down before you look at your pass rate.

2. **Count administrations per item, not per test.** Ask your platform for the administration count of your single most-used item over the last 12 months. Most teams have never asked. If one item has been seen by 8,000 applicants in a market where candidates share question screenshots, treat it as public.

3. **Set a retirement rule with a trigger, not a date.** Retire an item when it crosses a stated administration count, or when its proportion-correct moves more than a set amount from its calibration value without a change in the applicant pool. "We refresh annually" is not a rule; it is a hope.

4. **Monitor proportion-correct drift monthly.** This is the cheapest leak detector you have, and it is exactly what NCME recommendation 2 asks for. An item that gets easier while the rest of the form holds steady is the signal. Investigate the item, not the candidate.

5. **Separate the practice bank from the live bank completely.** Public practice items are a feature — they reduce anxiety and construct-irrelevant variance. They must never be scoreable items. One shared item between the two sets converts preparation into pre-knowledge.

6. **Assemble forms, do not just shuffle.** Randomising option order within a fixed question set defeats nothing: the answer circulates as the question text, not as "option C". Build multiple parallel forms from a stratified bank, and check score comparability across them before you compare candidates.

7. **Put the obligation in the vendor contract.** The 2024 Act makes service providers separately liable for leakage, with a ₹1 crore ceiling and a four-year bar, precisely because custody of content is a contractual fact. Your assessment contract should name who holds the bank, who may export it, what happens on a confirmed compromise, and who pays for replacement items.

Two design choices reduce the problem at source. Constructed-response and scenario-based items leak less usefully than multiple-choice, because a circulated answer key is far harder to reproduce under time pressure than a letter — AssessAll's AI-graded scenarios and integrity bands are built on that logic. And a bank large enough to rotate is the actual fix: if yours is too small to retire anything, commissioning additional calibrated items through a custom assessment service is cheaper than defending a loosened cut score.

## What item security does not fix

It does not establish who sat the test — that is identity assurance, and it is a separate control. It does not stop a candidate using an AI assistant in the moment. It will not rescue a test that was never validated for the role. And tighter exposure control costs something real: more items to write, slightly less measurement precision per item administered, and a longer test to recover it.

Those are manageable costs. A cut score that has drifted six points without anyone noticing is not.

## The takeaway

Proctoring protects the session; item security protects the score, and only one of them is usually budgeted. Count your exposure ratio and your most-administered item this week — if you cannot produce either number, your assessment is less secure than the public exams that got a new criminal statute written for them.

---

Source: https://www.assessall.com/blog/india-criminalised-question-paper-leaks-in-2024-your-hiring-assessment-still