The cognitive-safety moment India’s Air Force cannot postpone

Opinion  |  Aviation safety  |  Indian Air Force  |  Document no. SMF/OPN/IAF-SAF/2026/002  |  19 April 2026

The MiG-21 has been retired. A Tejas has fallen in Dubai. Between them lies a causal pattern the Indian Air Force has the evidence, the science and the institutions to address — if it chooses to.

By the Safety Matters Foundation — 19 April 2026

On 26 September 2025, at Chandigarh Air Force Station, the Indian Air Force retired its last MiG-21 Bison after sixty-two years of service. Air Chief Marshal A.P. Singh flew historic Bison CU2777 on the type’s final sorties. The ceremony drew the Defence Minister and the Chief of Defence Staff. Two months later, on 21 November 2025, Wing Commander Namansh Syal was killed when his HAL Tejas nosedived into the tarmac during a flying display at the Dubai Airshow. It was the first Tejas fatality, and the second Tejas hull loss in twenty months.

Between those two events sit ten years of accidents and near-misses we have compiled, cross-checked and counted: sixty-five documented incidents, ninety-three lives, the death of a Chief of Defence Staff in the Nilgiris, the first combat aircraft losses India has taken since Kargil, a friendly-fire shootdown of an Indian helicopter by an Indian surface-to-air missile, a fatal mid-air collision at the country’s premier airshow the day before it opened, the disappearance of a transport aircraft with twenty-nine aboard that was found only eight years later at three-and-a-half kilometres under the Bay of Bengal, and the quiet, cumulative attrition of fighter and trainer airframes that ministers patiently recite to Parliament each year as “accident rates” expressed in decimals.

The numbers have, in fairness, improved. In the Defence Ministry’s telling, the IAF’s accident rate has fallen from approximately 0.9 per 10,000 flight hours in 2020 to 0.2 per 10,000 by 2024. Converted to the international convention — per 100,000 hours — that is a move from roughly nine to two. It is real progress. The US Air Force, for comparison, reported 1.9 Class A mishaps per 100,000 hours in fiscal year 2024, a four-year high in its own trend line. On headline rate alone, the IAF has closed most of a historical gap.

So the question is no longer “is the IAF getting safer?” It is. The question is what — specifically, programmatically — remains to be done. And here the decade’s record tells a more instructive story than the rate line does.

What the cases have in common

On 19 February 2019, the day before Aero India 2019 opened at Yelahanka, two BAe Hawks of the Surya Kiran Aerobatic Team collided during a rehearsal. Wing Commander Sahil Gandhi was killed. The preliminary finding pointed to pilot error during a mirror manoeuvre — a classic skill-based error at the end of a high-tempo rehearsal week.

On 27 February 2019, within ten minutes of takeoff from Srinagar, an IAF Mi-17V-5 of 154 Helicopter Unit was hit by an IAF SPYDER surface-to-air missile. Six airmen and one civilian on the ground were killed. The helicopter’s IFF transponder had been switched off; under the tempo of the ongoing engagement with Pakistani fighter packages, the radar return was misclassified as a possible hostile drone. Air Chief Marshal R.K.S. Bhadauria subsequently called it “a big mistake.”

On 1 February 2019, two test pilots — Squadron Leaders Samir Abrol and Siddhartha Negi — had already been killed when an upgraded Mirage 2000 crashed on takeoff at HAL Bengaluru during a user acceptance flight. The Court of Inquiry pointed to a software or control anomaly tied to the HAL upgrade, compounded by the failure of a runway arrester barrier — a double-failure event whose causes span the HAL/IAF interface rather than sitting in either organisation alone.

And on 8 December 2021, a Mi-17V-5 carrying Chief of Defence Staff General Bipin Rawat, his wife, Brigadier L.S. Lidder and eleven others struck a hillside near Nanjappachatiram in the Nilgiris in degrading weather on a short transit to the Wellington staff college. Thirteen died at the scene. Group Captain Varun Singh — who a year earlier had won the Shaurya Chakra for recovering a stricken Tejas after a triple flight-control-channel failure — was the sole temporary survivor and died of his injuries a week later. The IAF’s Court of Inquiry, and a rare public finding by the parliamentary Standing Committee on Defence in December 2024, attributed the crash to Controlled Flight Into Terrain in adverse weather, with human factors at the crew level identified as the proximate cause.

These four cases — Yelahanka, Budgam, Bengaluru Mirage, Coonoor — are not mysterious. Each has a fingerprint that the published human-factors and cognitive-safety literature has been mapping for decades.

Yelahanka is skill-based error under time pressure: the sort of event every CRM module has covered since the 1980s, but which reads differently when you examine how rehearsal schedules, crew duty periods, and expectation violations interact in high-tempo demonstration flying.

Budgam is a level-two situational-awareness failure on the air-defence side: the operators perceived the return; the meaning was misclassified under combat stress, task saturation and fragmented IFF discipline. Contemporary cognitive-safety literature describes exactly this pattern under the heading of “fratricide prevention” and has operational countermeasures.

The Bengaluru Mirage crash sits in the literature on automation surprise and post-upgrade verification-and-validation: when software changes the aircraft’s behaviour in ways the pilot’s mental model hasn’t caught up with, and when an administrative barrier (the arrester) is the only remaining line of defence.

Coonoor is plan-continuation bias in its purest operational form: the NASA-Ames review of thirty-seven investigated accidents found approximately seventy-five per cent of tactical-decision errors were plan-continuation decisions — the unconscious drive to press on with the original plan rather than return or hold. Landman and colleagues’ 2022 review in Frontiers in Physiology shows that rotary-wing operations produce spatial-disorientation incidence roughly 5.7 times higher than fixed-wing. In adverse weather in mountainous terrain, the Coonoor profile is precisely where that risk concentrates.

What these fingerprints have in common is not that the aircraft were old, or that maintenance was at fault, or that equipment failed, though each of those stories is also present elsewhere in the record. The common thread is that the human operating system — attention, expectation, decision-making under stress, mode awareness, spatial orientation — failed predictably, in ways that the science has characterised and the civil airlines and several allied air forces have been engineering around for more than a decade.

That is what we mean by cognitive safety.

The shape of a programme

A formal cognitive-safety programme is not exotic. Its components exist — and crucially, the Indian components already exist. The Human Factors Analysis and Classification System developed by Shappell and Wiegmann, now a US Department of Defense standard at version 8.0, is the accident-coding taxonomy the IAF should adopt. Endsley’s three-level situational-awareness construct has simulator-ready measurement instruments. Unexpected-event simulator training for startle and surprise is embedded in the US Navy, USAF and several NATO operational conversion syllabi. Threat and Error Management is the sixth-generation Crew Resource Management framework, with plan-continuation bias and automation complacency as named threats. Fatigue Risk Management Systems are codified in ICAO Annex 6 and deployed in commercial aviation worldwide.

But here is the point worth making in India: the two most important pieces of the puzzle are already ours.

The first is pSuMEDhA — Psychomotor Evaluation Designed for Aviators — an indigenous psychometric cognitive test battery developed at the Institute of Aerospace Medicine in Bengaluru. Thambidurai, Sharma, Sowgandhi and Biswal published the first comparative validation of pSuMEDhA against the international benchmark CogScreen-AE in the Indian Journal of Aerospace Medicine in late 2024. Their fifty-subject comparison found that pSuMEDhA’s Digit Symbol Substitution Test and CogScreen-AE’s Symbol Digit Coding produced congruent speed and accuracy measures (Pearson’s r = 0.6, p = 0.000) — in other words, on the canonical working-memory task, India’s home-grown battery performs on par with the international standard. pSuMEDhA also includes a Threat Perception and Estimation Test that CogScreen-AE does not offer — a module that produces a threat-index score directly relevant to the combat-cognitive scenarios Budgam and Operation Sindoor exemplify. The authors explicitly flag that pSuMEDhA’s flight-simulator validation is the natural next step. That step is a matter of institutional decision, not of science.

The second is Perceptiva Chakshu Yan — a Safety Matters Foundation research construct authored by Capt. Amit Singh FRAeS. In its ideation form (Proposal 1024-01, October 2024) Chakshu Yan was a real-time eye-tracking and near-infrared spectroscopy pilot-state monitor aimed at fixation, cognitive overload, and inadequate situational awareness — the three causal fingerprints we have already seen in Air France 447, Colgan 3407, Eastern 401, TransAsia 235, and, in the Indian record, in the Coonoor and Yelahanka events. Its peer-review iteration (Indian Journal of Aerospace Medicine, manuscript IJASM_15_2026) develops Chakshu Yan into a full methodological framework: a comparative modality review, a 2,400-subject synthetic-population simulation, a twelve-item Cognitive Load and Overload Questionnaire with strong psychometric properties, and a fieldable multimodal sensing stack — pupillometry, blink/gaze dynamics, heart-rate variability, respiration, and ambient-light normalisation — achieving modelled classification performance of macro-F1 0.84 in the primary configuration and 0.87 with fNIRS validation.

The two instruments are complementary. pSuMEDhA is a psychometric assessment administered off-line, producing a profile of an aviator’s cognitive capabilities that belongs at the selection and periodic-evaluation tier. Chakshu Yan is a real-time multimodal monitor administered during task performance, producing continuous estimates of workload and overload risk that belong at the simulator and, where headgear is acceptable, the operational tier. Neither substitutes for the other. Together, they approximate the two-tier cognitive-safety architecture the international literature converges on.

We would therefore put it this way. The IAF should establish a Cognitive Safety Branch, sitting between the Directorate of Flight Safety and the Institute of Aerospace Medicine, with five core functions. First, adopt HFACS 8.0 as the standard IAF accident-coding taxonomy and publish anonymised annual distributions. Second, accelerate the flight-simulator validation of pSuMEDhA so its sub-scores — threat perception, vigilance, psychomotor tracking — become routine inputs into aircrew training emphasis and re-currency decisions. Third, pilot Chakshu Yan at the operational-monitoring tier, beginning at the Air Force Academy and selected Operational Conversion Units’ simulators, with cross-construct validation against pSuMEDhA. Fourth, embed unexpected-event simulator scenarios targeting startle and surprise into operational conversion, and make plan-continuation bias an explicit element of TEM-grade CRM. Fifth, build an IAF-specific Fatigue Risk Management System with circadian and sleep-homeostatic modelling and wearable fatigue monitoring for high-tempo windows.

None of this is revolutionary. None of it is expensive at the scale of a fighter-procurement programme. All of it is evidence-based. And the two centrepieces — pSuMEDhA and Chakshu Yan — are already Indian.

The transparency dividend

There is a second argument for a programme of this shape, which has less to do with training and more to do with governance. Over the last decade the public record of the IAF’s safety performance has improved in one respect — ministerial replies to Parliament have become more regular and more numerate — while remaining sparse in another: Court of Inquiry findings rarely enter the public domain. The December 2024 Standing Committee report on Coonoor, with its explicit human-factors attribution, is the exception that proves the rule. Outside that case, almost every row in our register ends with “CoI ordered” and very little more.

There is a serviceable argument for operational-security discretion on investigation detail, especially in combat or border-related cases. There is not, in our view, a serviceable argument against a USAF-style annual Safety Statistical Abstract that publishes Class A-equivalent rates per 100,000 hours, disaggregated by platform and phase of flight, alongside an HFACS distribution that shows — anonymously — what the service is seeing and what it is learning. The USAF publishes this. The US Naval Safety Command publishes this. The UK Ministry of Defence publishes a more restricted but still public annual health-and-safety statistics report. In 2026 there is no meaningful reason the IAF could not publish a version of the same.

What we are not saying

We are not saying the IAF is unusually unsafe. The decade’s own rate line argues the opposite. We are not saying human error is everywhere or that technical faults have ceased to matter; the Porbandar Dhruv swashplate and the Nashik Su-30 post-overhaul losses are real airworthiness stories that need their own institutional response. We are not saying that any one case — Coonoor, Budgam, Yelahanka — is reducible to a single cognitive pattern; real accidents are always layered, and the HFACS framework exists precisely to preserve that layering.

What we are saying is simpler. The causal patterns visible in the decade’s record are the patterns the cognitive-safety field has characterised in detail. The instruments that would convert those patterns into targeted interventions already exist — some in allied services and civil aviation, and, crucially, two of the most important in India itself. pSuMEDhA, validated in 2024, belongs at the selection and evaluation tier. Chakshu Yan, conceptualised in 2024 and methodologically developed in the 2026 IJASM draft, belongs at the operational-monitoring tier. Together with the institutional base — the Institute of Aerospace Medicine, the Directorate of Flight Safety, TACDE — and an evidence base of sixty-five documented incidents over ten years, the IAF has everything needed to operationalise a programme. The MiG-21 has been retired. A chapter has closed. The cognitive-safety chapter is ready to be opened.

About this opinion piece

This is a Safety Matters Foundation opinion publication, released alongside the Foundation’s research paper Indian Air Force Safety, 2016–2025 (document no. SMF/RES/IAF-SAF/2026/001). The Foundation is not affiliated with the Indian Air Force, the Ministry of Defence, Government of India, or any aircraft manufacturer. This piece expresses the considered view of the Foundation’s safety research programme and is based on open-source material. Any errors or omissions are the Foundation’s; the Foundation welcomes corrections and clarifications from the service, its institutions, and the broader research community.

Citation: Safety Matters Foundation. The cognitive-safety moment India’s Air Force cannot postpone. Opinion paper, SMF/OPN/IAF-SAF/2026/002, April 2026. Contact:Admin@safetymatters.co.in