Open record · 2004–2026

When the paper
gets out.

Between 2004 and 2026, at least 117 public examinations in India were publicly reported as compromised — question papers photographed in strongrooms, OMR sheets altered after the fact, paid impersonators sitting in place of candidates, and answer keys moving on WhatsApp hours before the bell.

This is the running record: one row per incident, every row linked to the report it came from. It counts institutions and outcomes, not people. Filter it, chart it, download it — and if a leak is missing, add it.

117 of 117 incidents

See all incidents →
Incidents
117
2004–2026
Officially confirmed
96
82% of the selection
Candidates affected
2.3crore
reported in 44 of 117 incidents
Exams voided
67
cancelled or re-held
Arrests recorded
1,862
across 68 incidents
Convictions recorded
114
in just 8 incidents

01 — The trend

Reported leaks have not tapered off. They have compounded.

Each column is a year; each block is one incident, shaded by how strong the evidence is. Click a year to filter the whole page to it.

Incidents per year, by strength of evidence

Colour runs dark to light with evidentiary strength. Grey sits outside the scale: officially investigated and rejected.

  • Confirmed
  • Alleged
  • Suspected
  • Denied

Bands mark the governing coalition at the time; the 2014 boundary falls in May, so that year straddles both. Counts reflect what has been reported and indexed, not everything that happened — see the caveat on the right.

Show the numbers
YearConfirmedAllegedSuspectedDeniedTotal
200410001
200610001
200810001
200910001
201010001
201150005
201280008
201350005
201411002
201530003
201620013
201741005
201850005
201930003
202040004
20211510016
20221520017
2023741012
2024721313
202540105
202632016

Rate per year

UPA (to May 2014)
24 incidents over 10.1 years
2.4/yr
NDA (May 2014 onwards)
93 incidents over 12.1 years
7.7/yr

Read this carefully

The rise here is partly real and partly an artefact. Recent leaks are covered by more outlets, indexed by search engines, and litigated in public — a 2007 state recruitment scandal reported once in a regional daily may simply never have reached this list. Treat the shape as the floor of the problem, not its measurement.

Only 8 of the 117 incidents in this record have a conviction attached to them. Sources are one report per row — pending cases stay uncounted

02 — Geography

A handful of states carry most of the record.

Click a bar to filter the page; click the state's name to open its own page. “All India” covers nationwide examinations — NEET, SSC, CBSE boards, railway recruitment — which affect candidates everywhere.

Incidents by state or territory

Top 15 of 26 places in the current selection.

  • Confirmed
  • Alleged
  • Suspected
  • Denied

+11 more with fewer incidents — see every state.

An incident spanning two states is counted in both, so the bars sum to more than the incident total. Nationwide exams appear only under “All India”.

Show the numbers
PlaceConfirmedAllegedSuspectedDeniedTotal
All India1211216
Madhya Pradesh1011113
Uttar Pradesh1210013
Rajasthan920011
Bihar61007
Haryana51006
Maharashtra51006
Uttarakhand60006
Gujarat41005
Jharkhand22015
Punjab20114
Assam30003
Chhattisgarh21003
Delhi21003
Himachal Pradesh30003
Jammu & Kashmir20002
Karnataka20002
Tamil Nadu20002
Andaman & Nicobar Islands10001
Andhra Pradesh10001
Arunachal Pradesh10001
Kerala10001
Manipur10001
Odisha10001
Telangana10001
West Bengal10001

Largest single incidents

By candidates affected, where reported.

  1. 1 48L
  2. 2 23L
  3. 3 22.7L
  4. 4 20L
  5. 5 15.6L
  6. 6 9.7L

Figures mix definitions — some sources count everyone registered for the exam, others only those whose sitting was voided. They are an order-of-magnitude signal, not a precise headcount.

03 — Institutions

Mostly state recruitment boards — but no tier is clean.

Central agencies run the largest examinations, so their failures affect the most people. State boards fail far more often.

94
State bodies
80% of incidents
23
Central bodies
20% of incidents

Examination bodies with the most incidents

Renamed institutions are merged — Vyapam, MPPEB and MP ESB are one body across three names.

  • Central body
  • State body

Jurisdiction is inferred from where each body's examinations were held. Click a bar to filter the page to that body.

Show the numbers
BodyTypeJurisdictionIncidents
Vyapam / MP ESBStateMadhya Pradesh12
CBSECentralNationwide9
NTACentralNationwide5
RSMSSBStateRajasthan4
UKSSSCStateUttarakhand4
BPSCStateBihar3
Indian Railways (RRB / RRC)CentralNationwide3
JSSCStateJharkhand3
RPSCStateRajasthan3
AIIMSCentralDelhi2
BFUHSStatePunjab2
BSER / RBSEStateRajasthan2
BSSCStateBihar2
DSSSBStateDelhi2
GSSSBStateGujarat2
HBSEStateHaryana2
HSSCStateHaryana2
MSCEStateMaharashtra2
SSC (central)CentralNationwide2
UKPSCStateUttarakhand2
Vyapam / MP ESB appears 12 times in this record — 12 separate examinations in Madhya Pradesh. The most-repeated body in the current selection

Why the imbalance? There are far more state recruitment exams than central ones, and state boards are typically smaller, more locally staffed and more exposed to political pressure. Several — Himachal's HPSSC, Uttarakhand's UKSSSC — were suspended, dissolved or had their functions transferred after the incidents listed here.

04 — Scale

One cancelled exam can void a year for millions.

A different measure from the one above: not how many exams failed, but how many people were sitting them.

Candidates affected per year

Summed across incidents where a source gives a figure. Blank years are years with no reported figure, not years with no incident.

This is deliberately a separate chart from the incident counts. The two measures share no axis and must not be read against one another. One entry alone, UP Police Constable Recruitment Exam 2024, accounts for 48L candidates, 21% of the total plotted here.

Show the numbers
YearCandidates affectedIncidents reporting a figure
20091141 of 1
201152,0002 of 5
20122861 of 8
201511,87,0003 of 3
201625,7521 of 3
20171,90,0002 of 5
20182,50,0001 of 5
201955,0001 of 3
202084,8003 of 4
202145,11,6416 of 16
202212,35,0005 of 17
202331,21,2066 of 12
202492,07,7697 of 13
20253,14,0003 of 5
202622,77,0002 of 6

05 — What happened next

Exams get cancelled. Cases rarely close.

The response to a leak is fast and visible; the consequences for those who caused it are slow and mostly invisible in the public record.

An FIR was filed or arrests were made 9682%
An official probe was opened (CBI, SIT, STF, ED) 6858%
The examination was cancelled 6051%
A retest was held 2521%
At least one conviction is recorded here 87%

This is the weakest number on the page. Each row carries one source, usually written soon after the leak; convictions arrive years later and are rarely back-filled. Read it as “convictions we can currently evidence”, not “convictions that exist”.

No action of any kind was reported 43%

Totals in the current selection

Arrests
1,862 across 68 incidents
Convictions
114 across 8 incidents
High confidence
87 of 117 rows

An arrest is not a conviction. 5 entries in the record ended with the allegation officially investigated and rejected, or with the “leaked” paper turning out to be fake: NEET-UG 2016 (Phase II), JPSC Combined Civil Services Preliminary Exam (11th-13th CSE), UGC-NET June 2024 and 2 more. Those rows are marked Denied, and they stay in the record because deleting a retracted scandal would quietly inflate the confirmed count.

06 — Deaths

The deaths reported alongside these scandals, with their caveats intact.

4 entries carry a reported death figure, ranging from 1 to 23. They are deliberately kept out of the summary figures at the top of this page, and they are never added to one another, because the numbers do not mean what a single headline figure would make them mean.

23 reported deaths

MP Pre-Medical Test (PMT) 2013

7 Jul 2013 · Madhya Pradesh

SCAM-WIDE, not this single exam: the MP SIT acknowledged 23 'unnatural' deaths of people linked to the Vyapam scam (~7 suicides, ~11 road accidents, ~3 alcohol-related illness); opposition/some press put it at 40-46. CBI probed 24 such deaths and ruled out foul play in every one it examined. Deaths span the whole 2009-2015 scam; attributed to the 2013 PMT row only for indexing. Mostly not suicides and highly contested. Sources: The News Minute, IBTimes, Business Standard.

1 reported deaths

RAS (Rajasthan Administrative Service) Preliminary Examination 2013

26 Oct 2013 · Rajasthan

Not an aspirant: the alleged mastermind of the leak died under suspicious circumstances during a custody transfer (Varanasi->Jaipur) in 2014; his family alleged murder. No suicide. Source: Careers360.

4 reported deaths

NEET-UG 2024

5 May 2024 · All India

>=4 NEET aspirants' suicides were reported by families/press as linked to the leak and exam cancellation (2 in Nagpur, incl. an 18-yr-old who left a note refusing the retest; cases in Kokrajhar, Assam and Kalaburagi, Karnataka). Causation (leak vs. general exam stress) is contested and not officially established; count is a floor of individually-reported cases. Sources: India.com, The Quint, Careers360.

21 reported deaths

NEET-UG 2026

3 May 2026 · All India

Press reports, student organizations and protest tallies at Jantar Mantar report up to 21 aspirant suicides across multiple states (including Rajasthan, Tamil Nadu, UP, Delhi, Karnataka) following the 12 May 2026 exam cancellation and re-test anxiety. This is a contested press/protest tally, not an official causal count, and conflates leak distress with general exam pressure. Sources: The Wire, The Indian Express, Al Jazeera, Hindustan Times.

Open data · CC BY 4.0

This record is incomplete. Help finish it.

Every row here is one news report away from being verifiable — and there are leaks we have certainly missed, especially state recruitment exams before 2015 and cases reported only in regional-language press. The dataset is one CSV file. Adding to it takes a pull request.

1 Find a source. A named news report, court order, official notice or RTI reply. One incident, one primary link.
2 Add one row to exam_leaks.csv. Leave a field blank rather than guessing it.
3 Open a pull request. A preview build validates the row; a maintainer checks the source against the claim.
4 Not a coder? Open an issue with the link and we will transcribe it for you.