Andreas Fankhauser
Lustenberger + Partners KLG, Zurich
fankhauser@lplegal.ch

Josephine Nalubega
Kampala
josephine_n@outlook.com

The problem

When a state investigates an air crash on its own territory, its report becomes the official record of what happened. Under current rules, that record stands whether or not it is complete, accurate or even credible. No one is charged with checking it. The downing of Ukraine International Airlines Flight PS752 made that gap impossible to ignore.

In March 2026 the International Civil Aviation Organization (ICAO) responded with Amendment 20 to Annex 13, the accident investigation annex to the Chicago Convention. It addresses the conflict of interest that arises when the State of Occurrence, which conducts the investigation, is also the state that shot the aircraft down. It does not address the report itself.

The case for a validation step has been made in the academic literature.^[1] This article makes the case from the practitioner’s side. It reads Iran’s final report against Annex 13 and asks why a flawed report nonetheless stands. It then uses the Swiss example to show that safety reports, although meant for prevention only, end up as evidence in court, and argues that the reform just adopted needs a final step: validation before publication.

What the report does – and does not – say

On 8 January 2020, a Boeing 737-800 of Ukraine International Airlines took off from Tehran for Kyiv and was shot down minutes later. All 176 people on board died. Most were Iranian nationals, many of them also Canadian citizens or residents.

Iran, as the State of Occurrence, investigated through its Aircraft Accident Investigation Board, which sits within the Civil Aviation Organization under a government ministry. Ukraine (State of Registry and of the Operator), the United States (State of Design and Manufacture) and France (engines) took part through accredited representatives. Canada, the United Kingdom, Sweden, Germany and Afghanistan were invited to appoint experts as states with a special interest. The ICAO, invited ‘to appoint a team of advisors to observe the process’, followed the investigation through a representative. The final report was issued on 15 March 2021.

The report’s account of the launch is brief. Early that morning a mobile air-defence unit was moved and not properly realigned, leaving a heading error of 105 degrees. When PS752 appeared, the operator saw a target on a bearing that did not match the aircraft’s true position. At 06:14:19 he attempted to report it to the coordination centre, but the message was never relayed. No command came. He fired at 06:14:39 and again at 06:15:09, and the aircraft crashed at 06:18:23.

The report names the missile launch as the cause. It lists one contributing factor: the mitigating measures in airspace risk management ‘proved to be ineffective due to the occurrence of an unanticipated error in threat identifications’. Its recommendations follow that framing – more rapid airspace warnings, more cautious risk assessment and better use of open-source information by airlines. None is addressed to the armed forces. The only one that even mentions them asks ICAO for guidance on cooperation between investigators and military authorities. It follows an admission: the team found it ‘challenging to investigate the actions and their root causes within the military sector’ and therefore asked the military itself to investigate the events leading to the launch.

Three features deserve a lawyer’s attention, and the report is open about all three. First, the account of the launch came from the party under examination: ‘The information in this section was provided by military authorities. The investigation team was able to see evidence that confirmed it.’ Second, the team ‘addressed three areas’ – military, civil and their interaction – but ‘the identification of the root causes and the provision of recommendations are confined solely to the civil area and its cooperation scope with the military one’. Third, the report avoids blame: the investigators say they used neutral wording and take no adversarial position towards anyone named.

The third feature is not a defect; it is what Annex 13 asks for. The problem is what the first two leave out.

Take the first feature. The team says it ‘was able to see evidence that confirmed’ the military’s account, but it never identifies that evidence. Elsewhere it lists, in general terms, the evidence it drew on – videos, recordings, photographs, interviews, judicial files, radar data and the flight recorders. Those can establish the time and place of the launches. They are unlikely to confirm what the report says happened inside the unit – the misalignment, the message that was not relayed, the absence of a firing command. On those points the reader is asked to take the military’s word.

The second feature makes this worse. The report does not ask why the operator’s message was not relayed, why he could fire without authority, how the unit was trained and supervised, or whether it was defensible to keep the airspace open to civil traffic during a military alert. These questions of command, control and airspace management go to the heart of prevention. Canada’s forensic team made the same criticism, and the report’s own description of its scope confirms it.

Procedural complaints add to the problem. Canada’s Special Advisor and the families’ association say that the crash site was cleared too quickly, that the flight recorders were only read out months later, and that victims’ devices were tampered with. If true, these allegations engage the Standards on the protection of evidence and the prompt read-out of recorders.

Annex 13 also requires an investigation authority independent of the state’s aviation authorities and of anyone else able to interfere with its objectivity. An authority inside the aviation regulator, under a ministry, working from facts supplied by the armed forces, sits uneasily with that requirement.

Canada had expert status only, with no accredited representative; it was not among the states consulted on the draft report and saw it only through Ukraine. Its Transportation Safety Board rejected the Iranian board’s position that military activities fall outside the scope of an Annex 13 investigation, and found its own questions only partly answered: ‘the report says what happened, but doesn’t address the why’.^[2] None of this, even if every allegation were confirmed, invalidates the report under current rules.

Why the report stands

Article 26 of the Chicago Convention obliges the state in whose territory an accident to a foreign aircraft occurs to hold an inquiry ‘in accordance, so far as its laws permit, with the procedure which may be recommended’ by ICAO. This recommended procedure is set out in Annex 13, which contains ICAO’s Standards and Recommended Practices for accident investigation. States may notify differences from those Standards under Article 38, but not – as the Annex itself records – from the three obligations Article 26 imposes directly. There are: to hold the inquiry; to admit the State of Registry's observers; and to send it the report and findings.

Annex 13 then allocates the roles. The States of Registry, of the Operator, of Design and of Manufacture take part through accredited representatives, with access to the evidence, the read-outs and the analysis meetings. A state whose citizens died may appoint an ‘expert’, with far more limited rights. Standard 6.3 requires the investigating state to send the draft report to the states taking part in the investigation and to invite their ‘significant and substantiated comments’. If it does not amend the report, it must append the comments at the commenting state’s request. The Standard says nothing about what happens if it does neither.

What happens is that the report is posted in ICAO’s library of final reports and stands as the official record. Posting is not endorsement, but there is no other record.

The precedents all point the same way. ICAO’s own report on Korean Air Lines Flight 007 (1983), though not an Annex 13 report, was reviewed by the Air Navigation Commission, which found it hard to validate conclusions that could not be fully explained because the Soviet Union had withheld the flight recorders. The Council adopted it anyway, and a US court later admitted it as trustworthy evidence. EgyptAir Flight 990 (1999) produced competing US and Egyptian reports, and the US one stood. The Ethiopian report on Flight ET302 (2022) was posted although the comments of the US and French investigation authorities had not been appended. PS752 followed the pattern.^[3]

No forum has ever ruled on the validity of a report with general effect. The Ontario court that heard the families’ claims against Ukraine International Airlines noted in passing that they had no reliable investigation report ‘as Iran’s report was not credible’ – a remark that binds no one. Article 84 of the Convention lets the ICAO Council decide disputes about the interpretation or application of the Convention and its Annexes, with an appeal to an arbitral tribunal or the International Court of Justice (ICJ). The Council’s 2025 decision holding Russia responsible for the downing of Flight MH17, now under appeal, shows that it is prepared to use that power, and a state could in principle take non-compliance with Article 26 and Annex 13 to the Council. None has.

The PS752 proceedings that Canada, Sweden, Ukraine and the United Kingdom brought before the ICAO Council in January 2024 are based on Article 3bis, the ban on the use of weapons against civil aircraft, not on the Article 26 duty to investigate. Their parallel case before the ICJ under the 1971 Montreal Convention concerns Iran’s duty to investigate a criminal offence and report its findings.^[4] The litigation targets the shooting-down, not the safety report.

Participating states can reject a report or publish their own analysis, but that only signals that they regard it as unreliable. It does not make it invalid.

Safety reports do not stay in their lane

The orthodox answer is that this is a problem for the safety community, not for lawyers. An Annex 13 report serves prevention, not liability. Liability is for courts and prosecutors with inquiries of their own, and Standard 5.12 restricts the disclosure of investigation records for other purposes, subject to a balancing test – though the final report itself is not among the records it protects. That describes the text, not the practice. In practice the safety report is often the central document in the liability lawsuits and criminal prosecutions that follow a crash.

Switzerland is a current example. It has long notified a difference to Standard 5.12: ‘Swiss legislation requires that all documents be made available to judicial authorities and aviation authorities.’ On that basis the Federal Criminal Court held in 2017 that the files and the final report of the Swiss Transportation Safety Investigation Board (STSB) are evidence in criminal proceedings, to be weighed freely by the court. Switzerland has since moved towards ‘just culture’: a revision of the Aviation Act now before Parliament would allow statements given to the STSB to be used in criminal proceedings only with the consent of the person who gave them, and would exempt slight negligence from punishment for disrupting public traffic where the person did everything reasonable to assist the safety investigation.^[5]

What the reform leaves alone is telling. The final report stays public, cannot be challenged – the Federal Administrative Court confirmed in 2025 that it is not a contestable ruling – and can still be used as evidence, where it is weighed freely. Swiss commentators have warned that the published report may in practice undercut the protections built around the statements that fed it. The pattern – and Amendment 20, as the next section shows, repeats it – is that the law protects the inputs and leaves the output in play.

The lesson for the validation debate is direct. A document that courts and prosecutors use to decide fault must be accurate – which is exactly what the prevention rationale, taken seriously, already demands. An inaccurate report does not stay in its lane. It is cited by scholars and journalists, it is used in court depending on the jurisdiction, and it becomes the history of how the victims of an air crash died.

Amendment 20: the right diagnosis, an incomplete cure

Reform began with a working paper tabled by Canada and a broad group of co-sponsors at the 2022 ICAO Assembly. Its diagnosis was conflict of interest: a State of Occurrence responsible for the downing investigates its own conduct, which puts independence, evidence, speed and transparency at risk.

The Council’s answer is Amendment 20 to Annex 13, adopted on 27 March 2026 and applicable from 23 November 2028. It adds guidance for investigations in which an actual or perceived conflict of interest could undermine confidence in the findings. It reminds the State of Occurrence that it may delegate the investigation to another state or to a regional investigation body – an option Annex 13 has long offered, and which Ukraine used for MH17. It encourages that state to invite ICAO or other states to observe and to give the public timely, verified factual information. The Council also made clear that investigators must have unrestricted access to all evidence without delay.^[6]

These are real and important improvements. They nonetheless fall short, for three reasons.

First, the language is permissive. Article 26 gives the investigation to the State of Occurrence. Taking it away from that state without its consent would require amending the Convention, not the Annex, so the Annex can only invite. The state that shot down the aircraft still decides whether to hand the investigation to someone else. It is not obvious why Iran, or any state in Iran’s position, would do so.

Second, PS752 shows that observers do not cure the defect. Iran invited ICAO to appoint advisers to observe the process. ICAO’s representative was kept informed, observed the recorder read-out and received the draft report for ‘advisory comments’. The report still rested on a military narrative, confined its root-cause analysis to the civil side and left Canada’s questions largely unanswered.

Third, and most importantly, the amendment addresses the input, not the output. Nothing in it changes what happens when the report is released: it is posted, it becomes the record, and comments that were ignored stay ignored. Delegation would not change that either. It would move the investigation to a state with interests of its own: a State of Manufacture may be suspected of shading a report to protect its aircraft, a State of Registry of protecting its pilots, as the EgyptAir dispute illustrates.

The missing step: validation before publication

The gap is at the end of the process, and that is where it should be closed. Adejoke Adediran has proposed a validation step written into Annex 13. Among current proposals, hers comes closest to doing so.^[7] The draft report, with the participating states’ comments, would go to ICAO as well as to the parties, and the finished report would be reviewed before release. The natural reviewer is the Air Navigation Commission. It already reviews safety recommendations addressed to ICAO. In the Korean Air Lines case it reviewed ICAO’s own report for compliance with ICAO provisions and the soundness of its conclusions – the right kind of review, though not of an Annex 13 report.

Domestic law knows the technique. Swiss law makes the STSB’s final report immune from appeal, but compensates in two ways: those involved must be consulted on the draft, and a report containing safety recommendations must be approved by the Board’s own commission of independent experts, which sits above the investigation service. That is an internal review before release in place of a remedy after it.

The review would ask two questions. Was the investigation conducted in compliance with Annex 13 – independence of the authority, protection and custody of evidence, timely read-out of the recorders, notification, and the handling of comments under Standard 6.3? And, is the report accurate – does it document the systemic and contributing factors, or does it stop at one operator’s error?

Non-compliance would block validation only where it affected the findings: a delayed read-out that distorted the evidence or a lack of independence that produced wrong conclusions, not a technical lapse with no bearing on the result. Failure to address a participating state’s substantiated comments – what happened to the US and French authorities on ET302 – would be a ground for refusal in itself. It would help a state in Canada’s position only if the right to comment were extended to states with a special interest, as Canada has urged.

The objections should be stated fairly. Sovereignty: states may see ICAO review of their investigations as external judgment. But ICAO already audits states’ safety oversight, and a validation review would assess the integrity of the investigation, not assign liability. The no-blame principle: the Swiss example answers it. A report that tribunals use to decide fault cannot claim immunity from scrutiny on the ground that it was never meant for that purpose. Politics and capacity: states may be reluctant to give ICAO a power that could one day be turned on them, and the Commission would need the resources to do the work. The first is a real obstacle, not a doctrinal one, and the counterweight is states’ own repeated commitment to safety and security. The second is a matter of design.

Two design choices would help with both. The review could be triggered only where a conflict of interest within the meaning of Amendment 20 exists or where a participating state has told ICAO that its substantiated comments were not appended. And, as a first step needing no amendment at all, ICAO could decline to post a report to which such comments have not been attached – a signal rather than a sanction, but a signal ICAO can send today.

Conclusion

PS752 exposed not so much the failure of one state’s investigation as a failure of the system. A state that causes an accident on its own territory remains the unchallenged author of the official record. Amendment 20, for all its merit, leaves that unchanged: it asks the author to behave better and offers help, but it does not check the result. A validation step would not have stopped the missile. It might have produced a report that explained how to stop the next one – which is the only thing Annex 13 was ever for.

Notes

^[1] Adejoke O Adediran, ‘Need for Validation of Annex 13 Reports’ (2026) 51 Air and Space Law 55. This article draws on her analysis where indicated.

^[2] Iran Aircraft Accident Investigation Board, Final Report (2021), 11, 14–17, 74–77 and 134–140; Government of Canada, The Long Road to Transparency, Accountability and Justice (2020) and Factual Analysis (2021); Transportation Safety Board of Canada, statement of 18 March 2021.

^[3] See Adediran, n 1, 61–63 and 72–73.

^[4] S v Ukraine International Airlines 2024 ONSC 3303, para 454, affirmed 2025 ONCA 587; ICJ cases 190 (Montreal Convention), 198 (Iran’s appeal) and 201 (Russia’s MH17 appeal).

^[5] Federal Criminal Court, TPF 2017 15; Federal Administrative Court, A-5346/2022 (2025); Federal Council dispatch of 12 November 2025, BBl 2025 3532; Ronny Fischer, AJP 2026, 182 and 899.

^[6] A41-WP/193 (Canada, 2022); ICAO news release of 27 March 2026.

^[7] See Adediran, n 1, 75–81, on which this section and the previous one draw.