The Wrong Box

Opinion  ·  Regulation & Human Factors

India breath-tests its pilots before every departure and drug-tests them roughly once a decade. When the two regimes finally met in the same occurrence this August, the occurrence itself was filed under the wrong heading. Both facts point in the same direction.

An occurrence, and two questions

On 4 August 2026, Air India flight AI2379, an Airbus A320 registered VT-EXO, lost approximately 300 feet of altitude in cruise between Phuket and Delhi. The aircraft carried 137 passengers and eight crew. Seventeen people were injured. Thirteen passengers and four cabin crew were admitted to hospital. The treating emergency department reported compression-type injuries and a spinal fracture in one of the cabin crew. The commander reported a severe injury to a crew member. The aircraft flew on for another hour and landed safely, its cabin crew working through their own injuries.

Two administrative acts followed. Both flight crew underwent psychoactive substance screening under standard procedure. And the occurrence was classified a Serious Incident.

Each of those acts opens a question. The first asks whether India’s substance-testing architecture is fit for the purpose it claims. The second asks whether India’s occurrence classification is being applied to the facts, or to something else. Taken together they describe a regulator that has built impressive machinery for the problem it finds easiest to see, and looked away from almost everything around it.

Where this came from, and what ICAO actually asked for

None of this began in Delhi, and India has never suggested otherwise.

The 27th ICAO Assembly adopted Resolution A27-12, urging the Council to give priority to effective measures for preventing and eliminating substance abuse among crew members, air traffic controllers, mechanics and other civil aviation personnel. The 29th Assembly followed with A29-16, on the role of ICAO in the prevention of substance abuse in the workplace. In 1995 the Council delivered the instrument: the Manual on Prevention of Problematic Use of Substances in the Aviation Workplace, Doc 9654.

India’s own account of that manual, given to the 42nd Assembly last year, sets out its contents in order: education in the aviation workplace, protocols for identification, approaches to treatment and rehabilitation, the employment consequences, and then a biochemical testing programme specifying testing procedures and drug detection periods.

Note the order. Education first. Identification and treatment second. Testing last. ICAO built a prevention architecture and placed the laboratory at the end of it, as the instrument of last resort rather than the load-bearing wall. India has built the laboratory and very little else.

Then, in 2001, the 33rd Assembly adopted the resolution that matters most to this argument. Its title is not incidental: Harmonization of drug and alcohol testing programmes. Its preamble records a continuing need for a consistent policy on implementation by national authorities of regulations concerning prevention and enforcement on the abuse of alcohol and drugs by personnel in safety-related occupations. Its operative clauses direct the Council to review existing guidance so that States can develop consistent prevention and testing programmes, and to develop the provisions necessary to achieve consistency among Contracting States’ testing programmes and their enforcement. States themselves are merely encouraged to foster consistency — the weakest verb in the Assembly’s vocabulary.

And here the argument stops being mine. India’s own working paper to the 42nd Assembly summarises A33-12 as directing the Council to develop provisions ensuring consistent alcohol and drug prevention and testing programmes across Contracting States, emphasising uniform enforcement, and urging States to align their efforts through harmonized substance abuse policies. Alcohol and drug, in one compound phrase, as a single object of policy. That is India’s reading of the resolution, filed at Montreal in July 2025 over the Government’s own name.

The paper goes further than the resolution does. A33-12 encourages States to foster consistency; India’s summary says it urges them to align. Whichever verb one prefers, India has told the Assembly what the obligation is. It has not told anyone why its own two programmes do not meet it.

Two consequences follow, and they point in opposite directions. The first is ICAO’s to answer: twenty-five years on, the direction to the Council remains unexecuted. Doc 9654 is still the first edition of 1995. No consistency provisions were ever developed. The second is India’s, and it is the one nobody has raised. A33-12 does not treat alcohol and drugs as two subjects requiring two policies. It treats them as one subject requiring one coherent approach, and it says so in its title. Harmonisation is the thing ICAO has been asking for since 2001. Domestically, it is precisely the thing India has not done.

Two substances, one Standard, two standards

Annex 1 to the Chicago Convention does not distinguish between alcohol and other drugs. Paragraph 1.2.7.1 bars a licence holder from exercising privileges under the influence of any psychoactive substance. Paragraph 1.2.7.2 bars problematic use of substances outright. ICAO’s own Manual of Civil Aviation Medicine places alcohol first in its list of psychoactive substances. One clause, one class of harm.

India has built two entirely different regimes underneath that single clause.

Nothing in ICAO’s framework asks for this split. It is a domestic invention — and it sits awkwardly against the Government’s own account of where the drug regime came from. Answering a starred question in the Lok Sabha in March 2020, the Ministry described the draft Part V as issued under Rule 24 of the Aircraft Rules 1937 and expressly in line with provisions of ICAO Annex-1. The same Annex. The same clause that governs alcohol. Two regimes, one parent, and no stated reason for the divergence.

The arithmetic of ten per cent

A line pilot flying a normal roster submits to something in the order of five hundred breath tests a year. Under Part V, the same pilot faces a one-in-ten annual chance of a urine sample being requested — approximately one test per decade.

Deterrence is a function of perceived probability of detection, not of the severity of the sanction attached to it. At one test per decade, delivered after the flight has already landed, the deterrent value of Part V is close to indistinguishable from zero.

The comparison in the record is stark. Answering Parliament in March 2020, the Ministry reported breath-analyser violations under Part III of 254 in 2017, 256 in 2018 and 244 in 2019, with a further 48 in the first two months of 2020 — a stable rate of roughly a quarter of a thousand a year, sustained across three years. The separate CAR covering air traffic controllers, dispatchers, maintenance, aerodrome, fire and rescue, vehicle and ground handling personnel produced 97 violations in 2019 and 81 in two months of 2020. Against those figures, the psychoactive substance programme produced four pilots and one air traffic controller in its opening seven months of operation.

One set of numbers reflects prevalence. The other reflects a sampling rate. It is not obvious which is which — and that is precisely the problem. The drug programme generates too little data to tell us anything reliable about the population it purports to monitor. It is not a surveillance system. It is a gesture in the shape of one.

A test that cannot measure the thing that matters

A breath analyser answers the operationally decisive question: is this person impaired now, at the gate, before the doors close. Urinalysis answers a historical one: did this person consume something, at some point in a window that for cannabis metabolites can stretch to weeks.

Those are different questions, and Annex 1 asks both — 1.2.7.1 concerns being under the influence, 1.2.7.2 concerns problematic use as a pattern. India’s drug regime reaches only the second limb, and reaches it through a proxy that cannot distinguish a dependent user from someone who made a single poor decision on a long layover three weeks ago. Meanwhile the first limb — impairment on the flight deck by anything other than alcohol — is not tested for at all, because the only test India runs is administered after the aircraft has parked.

Revision 5 to Part III, issued in February 2026, tightened the alcohol regime considerably: fuel-cell analysers, ISO-certified calibration agencies, camera recording of every examination preserved for six months, reporting through eGCA within 24 hours, post-flight testing of international arrivals within two hours of chocks-on, daily testing of 40 per cent of student pilots. This is serious regulatory craftsmanship. It is also, structurally, the regulator polishing the instrument it already owns rather than acquiring the one it lacks.

A threshold problem the regulator has already admitted

There is a further difficulty, and it comes from India’s own submission to ICAO rather than from any critic.

A drug test result is meaningless without a cut-off value: the concentration above which a sample is declared non-negative. Set it too low and you catch poppy seeds and prescribed codeine; set it too high and you catch nobody. It is the single most consequential technical parameter in the entire regime, and it determines whether a pilot keeps a licence.

According to the working paper, DGCA constituted its expert committee to establish those reference cut-off values in 2022 — the same year the CAR came into force on 31 January. The committee reviewed and set cut-offs for six substance groups: amphetamine and amphetamine-type stimulants, opiates and metabolites, cannabis as THC, cocaine, barbiturates, and benzodiazepine.

The Part V panel lists thirteen substances. Buprenorphine and norbuprenorphine, methadone, oxycodone, phencyclidine and propoxyphene do not appear in that list of six at all. Methamphetamine and MDMA might arguably be swept inside “amphetamine and amphetamine-type stimulants”, though the paper does not say so. On the face of India’s own account, the regime tests for substances against which no nationally determined threshold has been described.

And then, in the same document, India asks ICAO to fix exactly this: the paper observes that Doc 9654 offers limited guidance on a standardised methodology for establishing benchmark values, and that the omission leaves States without a clear framework for evaluating positive detections. The diagnosis is correct. The patient is also the complainant.

The ladder that runs backwards

Here the inconsistency becomes hard to defend on any principle at all.

A pilot who fails a first breath test loses their licence for three months. A pilot who fails a first confirmatory drug test loses nothing — they are referred for de-addiction and return to line flying after a negative retest, psychiatric clearance and a fitness certificate.

The therapeutic route is the one ICAO actually recommends. Annex 1’s 1.2.7.3 is built on identification, removal, treatment and return to duty. India has applied that model to the substances its own evidence places second, and a bare deterrent model to the substance the same evidence places first — the working paper filed at Montreal states plainly that, on the Ministry of Social Justice and Empowerment’s February 2019 findings, cannabis and opioids are the next most commonly used psychoactive substances after alcohol. The regulator has, in effect, decided that cannabis dependency is a health condition and a glass of wine eleven hours before report is a disciplinary offence. Both propositions may be arguable. They cannot both be arguable at the same time, from the same evidence base, in the same regulator.

There is a further wrinkle in that ladder worth putting on the record. The draft placed before Parliament in March 2020 was harsher than the instrument eventually adopted: it provided that a person who tested positive on a second confirmatory test would have their licence cancelled. By the time Part V took effect in January 2022, cancellation on the second occasion had become a three-year suspension, with cancellation pushed to a third. The regime was quietly softened between draft and adoption, and no public reasoning accompanied the change. A regulator entitled to make that judgement is also obliged to explain it.

The safe harbour that does not exist

Part V encourages organisations to promote voluntary self-declaration, with rehabilitation before return to duty. It is the most valuable provision in the entire instrument, and it is inert, because of what sits beside it in the statute book.

Most of the substances on the Part V panel are controlled under the Narcotic Drugs and Psychotropic Substances Act 1985. Consumption is an offence under Section 27, punishable by up to a year’s imprisonment for the specified drugs and six months for the rest. Section 64A grants immunity from prosecution to an addict who voluntarily seeks and completes de-addiction treatment at a recognised institution.

The CAR does not invoke Section 64A. It does not incorporate it. It does not create any equivalent. It does not oblige DGCA or the operator to refrain from referring a result to law enforcement, nor does it oblige them to refer. What it does do is require that a positive confirmatory sample be preserved for twelve months with full chain of custody.

So the pilot who self-declares, or who accepts the rehabilitation route after a first positive, generates court-quality documentary evidence of a criminal offence, hands it to a State agency, and receives in return no statutory protection whatsoever — only the unstructured discretion of officials to leave it alone. No rational person makes that trade. The provision that could do the most good in the entire regime is the one nobody will ever use, and the reason is a drafting omission that could be cured in a paragraph.

And nothing where peer support should be

After Germanwings 9525, Europe reached a conclusion that the aviation medical community had been urging for two decades: you do not find impaired colleagues by testing for them. You find them because someone who flies with them says something, and says it into a channel that will not destroy them. Regulation (EU) 2018/1042 made crew support programmes a requirement, alongside random alcohol testing and systematic psychoactive substance testing by operators — a three-legged structure.

India has two of those legs, one of them vestigial, and no support programme mandated anywhere in the CAR architecture. Testing without a support programme is not a safety system; it is surveillance with a medical vocabulary. It detects only the person who has already run out of ways to conceal, which in a dependency is very late indeed.

Every hour of every day, the colleague sitting three feet away holds better information about a crew member’s condition than any laboratory will ever produce. India has built no channel through which that person can act on it without ending a career — their own or someone else’s. The self-declaration clause in Part V is that channel on paper. We have already seen why nobody will walk down it.

And then there is the box

Which brings us back to AI2379, and to the second administrative act.

Annex 13 defines a serious injury by several alternative limbs. Hospitalisation for more than 48 hours commencing within seven days is one. A fracture of any bone other than a simple fracture of a finger, toe or nose is another. Either one suffices.

Annex 13 defines an accident as an occurrence between boarding and disembarkation in which a person is seriously injured as a result of being in the aircraft. No damage to the aircraft is required. And Annex 13 is explicit that the distinction between an accident and a serious incident lies only in the result — a serious incident is one where an accident was probable but did not happen.

On the facts publicly reported by the treating hospital on the first day, a member of the cabin crew of AI2379 sustained a spinal fracture. Four crew remained under medical care after all thirteen hospitalised passengers had been discharged. On either limb, the threshold is met. The result is in. This is an accident.

It is not a novel or contested reading. Cabin injuries in turbulence, with an undamaged aircraft that lands normally, are the textbook accident-without-damage. Singapore classified SQ321 in May 2024 on precisely this basis. India’s own Aircraft (Investigation of Accidents and Incidents) Rules 2017 carry the Annex 13 definitions.

Classification is provisional, and AAIB can revise it as the medical picture is confirmed. It should. Because the consequences are not cosmetic: the investigation and final-reporting obligations differ, the ICAO notification pathway differs, and India’s published accident count for 2026 is understated by one. Over a decade, a systematic tendency to file accidents as serious incidents produces a national safety record that flatters the very oversight system that generated it.

The pattern, and what it is made of

A regulator confronted with a classification choice will, unless something forces otherwise, choose the lighter procedural regime. That is not corruption. It is gravity. The counterweights are supposed to be published criteria, independent investigative machinery, and external scrutiny.

India is thin on all three. The Enforcement Policy and Procedures Manual, which governs whether a lapse produces a warning, a suspension, a cancellation or a financial penalty, is not published. In 2022, DGCA took 305 enforcement actions and levied ₹1.97 crore in 39 cases — and no external party can assess whether that pattern was consistent, because the yardstick is not in the public domain.

Substance-testing figures do reach Parliament. Ministers have answered on confirmatory positives under the psychoactive substance regime and on the action taken in each case, and that disclosure is real and ought to be acknowledged. But an outcome disclosed without its denominator cannot be audited. Parliament is told how many people were caught. It is not routinely told how many were tested, by which operator, measured against the ten per cent each is obliged to cover — and therefore cannot distinguish between a clean profession and a programme that is simply not running at the rate the CAR prescribes. Those two conditions produce identical headline numbers and opposite policy conclusions. Publishing the denominator, per operator, per year, would cost nothing and would settle the question in a single table.

What that opacity costs became visible in September 2017, when DGCA identified 132 Air India pilots and 434 cabin crew as having skipped mandatory pre- and post-flight breath tests, and proposed three-month suspensions across the group. Air India’s reply was not a plea in mitigation. Its chairman wrote that there had been no violation of any nature, and that the airline had not introduced medical checks at transit stations because of how it read the language of the CAR. The pilots’ union alleged selective enforcement, asking why only two narrow-body services had been picked out and wide-body flights left alone.

Set aside who was right. The instructive fact is that 566 crew came within reach of grounding over a disagreement about what a rule meant — and that the regulator subsequently clarified the CAR to state expressly that a crew member who enters the terminal building at a transit airport must undergo pre-flight breath testing. The drafting was fixed, quietly, after the enforcement crisis rather than before it. And because the enforcement yardstick has never been published, the selectivity allegation could be made and could not be answered. That is the structural cost of an unpublished manual: not that the regulator is necessarily acting inconsistently, but that it has deprived itself of any means of demonstrating that it is not.

There is a particular irony in this, and the working paper embodies it twice over. India asked ICAO to modernise Doc 9654 while operating a regime whose own coverage its paper cannot describe consistently: one paragraph says the CAR merely encourages testing of maintenance engineers, certifying staff, trainee pilots and instructors, while a later paragraph says organisations shall ensure they are tested. Cabin crew — who are covered by the alcohol CAR, and four of whom were injured on AI2379 — appear in neither formulation.

A State that asks ICAO to modernise, while its own instrument tests one in ten pilots after the fact, excludes cabin crew, offers a rehabilitation pathway nobody dares use, mandates no peer support, and publishes neither its enforcement yardstick nor its testing denominator, is asking the right question of the wrong body.

What reform actually looks like

None of the following requires primary legislation. Most of it requires a revision to a CAR and one clarificatory notification.

1. 01Create a statutory safe harbour. A notification extending Section 64A protection, or an equivalent non-referral undertaking, to any aviation professional who self-declares or completes rehabilitation under Part V. Without this, everything else is decoration.
2. 02Mandate peer support programmes as a condition of the operator’s permit, with confidentiality protections that survive contact with the regulator, on the EASA model.
3. 03Raise Part V coverage and move part of it forward. Ten per cent annually, post-flight only, is not a programme. A meaningful share should be unannounced and pre-duty, using oral fluid where impairment rather than history is the question.
4. 04Extend Part V to cabin crew, AMEs and certifying staff as a mandate, not an encouragement. Part III already covers cabin crew; there is no coherent reason Part V does not.
5. 05Harmonise the two ladders, as A33-12 asked. ICAO named the deliverable in 2001 and put it in the title of the resolution. If treatment before sanction is right for one psychoactive substance, it is right for the other, subject to the same safeguards — and if it is not, the regulator should say why in writing.
6. 06Publish cut-off values for the whole panel. Thirteen substances are tested; six substance groups have publicly described thresholds. A licence should not turn on a number the licence holder cannot see, and no result is defensible without a validated threshold behind it.
7. 07Publish the EPPM, and publish the denominator. Positive counts alone tell Parliament nothing it can act on. Publish tests conducted per operator per year against the mandated ten per cent, alongside outcomes, for both regimes. A yardstick nobody can see cannot be shown to have been applied consistently, and a catch rate without a test rate cannot be interpreted at all.
8. 08Publish occurrence classification criteria and reasons. Where an occurrence involves reported injuries, the classification decision should record which Annex 13 limbs were considered and on what medical evidence, and should be revisable on the record.
9. 09Take the A42 working paper seriously at home. India asked ICAO to update Doc 9654 and to supply a methodology for benchmark values. It should demonstrate the case by fixing its own instrument first — and then offer that instrument as the model.

Coda

Four cabin crew were injured on AI2379, one of them with a fractured spine, while doing their jobs in a cabin that had become briefly ungovernable. They then worked for another hour. Whatever the laboratory eventually reports about the flight deck, those four people are the moral centre of this occurrence, and the system owes them an accurate account of what happened to them.

An accurate account begins with the right word on the form. You cannot test your way to a safety culture — and you certainly cannot classify your way out of an accident.

Sources. ICAO Annex 1 (1.2.7.1–1.2.7.3); Annex 13 definitions; Doc 8984; Doc 9654 (1st edn, 1995); ICAO Assembly Resolutions A27-12, A29-16 and A33-12 (Harmonization of drug and alcohol testing programmes, 2001); ICAO A42-WP/214, TE/79, Guidance framework for evaluating psychoactive substance use among aviation personnel, presented by India, 29 July 2025; DGCA CAR Section 5 Series F Part III (Rev 5, 9 February 2026) and Part V (Issue I, 27 September 2021, effective 31 January 2022); Aircraft (Investigation of Accidents and Incidents) Rules 2017; Bharatiya Vayuyan Adhiniyam 2024; NDPS Act 1985, ss. 27 and 64A; Regulation (EU) 2018/1042; Lok Sabha Starred Question No. 271 of 12 March 2020 (Random Drug Testing); Ministry of Civil Aviation statement of 9 August 2026 on AI2379; contemporaneous press reporting.

The views expressed are the author’s own and do not represent any employer or institution.