The State’s Recording, the Court’s Reasoning, and the Unfinished Test of Justice

A Bengal Gazette Investigative Legal Analysis

Bangladesh’s former prime minister was sentenced to death after a trial that relied on intercepted conversations, eyewitnesses, forensic examinations and evidence of killings during the 2024 uprising. The recordings have substantial corroboration. But the law asks questions beyond whether a voice is genuine: how evidence was obtained, what the science actually established, how individual criminal responsibility was inferred, and whether the defence had a meaningful opportunity to challenge that inference. A comparison of four tribunal judgments reveals why those distinctions matter.

---

The Objection That Came Too Late

On 21 October 2025, with Bangladesh’s International Crimes Tribunal approaching the end of a trial that would condemn a former prime minister to death, the lawyer appointed to defend Sheikh Hasina returned to the telephone recordings at the centre of the prosecution’s case.

The Criminal Investigation Department had examined the recordings. But Md Amir Hossain wanted an independent international examination. A tribunal judge asked why he had not raised the issue while questioning the forensic witnesses.

The lawyer acknowledged the omission.

“I could have mentioned it. I didn’t.”

The tribunal chairman declined to order another examination. Earlier in the exchange, a judge had pointed to examinations reported by the BBC and Al Jazeera.

It was a brief confrontation in a case overshadowed by the extraordinary violence of July and August 2024. Yet it exposed a question with consequences extending far beyond one defendant: when a lawyer fails to scrutinise an important piece of evidence adequately, what responsibility remains with the court?

The difficulty was not simply that the defence had raised its objection late. Under the tribunal’s governing law, evidentiary objections ordinarily had to be raised when evidence was submitted, with a limited exception for matters discovered later. But the same law empowered the court to consider admissibility on its own initiative and required reasons for evidentiary rulings.

Those provisions were already in force when the lawyer made his request.

On 17 November 2025, the tribunal convicted Hasina of crimes against humanity and sentenced her to imprisonment until natural death under the first charge and to death under the second and third charges. Former home minister Asaduzzaman Khan Kamal received the same two categories of punishment. Former police chief Chowdhury Abdullah Al-Mamun, who had become a prosecution witness, was sentenced to five years.

The judgments followed a national catastrophe. A United Nations fact-finding investigation estimated that as many as 1,400 people may have been killed between 15 July and 5 August 2024, most of them by security forces. Its findings described a systematic campaign of violent repression involving the former government and security apparatus.

The deaths, the injuries and the suffering of families are not speculative matters to be dismissed by disputing a particular recording. Nor does an allegation of procedural unfairness erase the prosecution’s substantial evidence.

But establishing that atrocities occurred and establishing the criminal responsibility of a particular defendant are different judicial tasks.

That distinction is indispensable when the punishment is death.

The State That Recorded Its Own Prime Minister

The telephone evidence has an institutional history that begins before the tribunal received a single recording.

Bangladesh’s National Telecommunication Monitoring Centre, or NTMC, was part of a surveillance architecture developed under successive governments, including Hasina’s own administration. Its capabilities, oversight and procurement history were examined in Bengal Gazette’s earlier investigation into the surveillance state.

That investigation asked a question familiar to democratic accountability: Who watches the watchers?

Hasina’s trial gave the question an unexpected consequence. Recordings associated with the state surveillance system that operated during her government later became evidence against her.

There is no inherent contradiction in that development. Evidence lawfully collected under one administration may properly be used to investigate wrongdoing by officials of that administration. Public authority must remain accountable even when power changes hands.

But the institutional reversal does not settle the legality of the original surveillance.

According to the testimony of special investigating officer Tanvir Hasan Zoha, identified as PW-53, investigators seized 69 audio clips and call-detail records relating to three telephone numbers from NTMC on 23 December 2024. Review began the following day. Selected conversations were later examined and introduced through the tribunal’s evidentiary process.

The distinction between collection and interception is essential. An investigator’s lawful seizure of an existing recording does not, by itself, establish the legal authority under which the original conversation was intercepted.

That does not mean the interception was unlawful. The telecommunications legislation operating in 2024 provided a statutory framework for authorised interception in specified circumstances involving national security and public order.

The unresolved documentary question is more precise: what authorisation applied to these particular calls, to which numbers, for what period, and under whose authority?

A second question arises from the selection of evidence.

Five conversations were identified in the investigator’s account as having been presented on three CDs. Selecting relevant material from a larger collection is ordinary prosecutorial practice. It becomes an evidentiary concern only if the selection removes material context, conceals relevant exculpatory information or prevents an adequate assessment of the recordings that were submitted.

The prosecution was not automatically obliged to play all 69 conversations in public. Some could contain irrelevant or sensitive communications involving third parties.

But the tribunal law contained a specific disclosure obligation. Section 12(2), introduced through the November 2024 amendment, required prosecutors to disclose, as soon as practicable, evidence in their possession that they believed might show innocence, mitigate guilt or affect the credibility of prosecution evidence.

The decisive issue is therefore not merely why five conversations were selected.

It is whether the available record establishes that the defence received the relevant context and material needed to test those five conversations.

Forensic authenticity begins with the file that an examiner receives. Evidentiary accountability begins earlier, with the history of how that file came into existence and reached the examiner.

What the Science Proved—and What It Did Not

The strongest argument against dismissing the recordings is that their authenticity was investigated.

The BBC’s 2025 investigation commissioned an independent examination by the audio-forensic organisation Earshot. That examination found no evidence of artificial synthesis or editing in the leaked recording it analysed.

The examination also suggested that the call may have been played through a loudspeaker and recorded by another device.

Those findings provide meaningful support for the authenticity of the material examined by Earshot.

They do not establish that the BBC’s leaked file was digitally identical to the files obtained from NTMC and examined for the tribunal.

That relationship needs to be demonstrated, not assumed.

An original interception recording, an exported file, a recording made from a loudspeaker and a later copy may contain the same conversation without being the same forensic exhibit.

Each has its own provenance. Each may have a different creation history, encoding, metadata and chain of custody. A test of one version cannot automatically authenticate every other version.

The distinction becomes especially important where a court treats the results of separate examinations as mutually reinforcing.

The Criminal Investigation Department’s work presents a related, but distinct, question.

The tribunal’s written judgment records the testimony of CID voice examiner Shahad Jubayer Lawrence, PW-45. In his examination of the Sheikh Hasina–Maksud Kamal conversation, the reported likelihood ratios for the compared voices were approximately 2.75 and 4.64.

Those figures have a specific scientific meaning.

A likelihood ratio compares how probable particular observations are under competing propositions. It is not the probability that the speaker is the alleged person. Nor is a ratio greater than one equivalent to identification beyond doubt.

The distinction is not a journalistic invention. The Forensic Science Regulator of England and Wales explains it in its guidance on interpretation and communication of forensic evidence.

Under that guidance’s illustrative scale, a likelihood ratio in the region of one to three may not materially distinguish between competing propositions, while a ratio between three and ten provides only slight support for one proposition over the other. The guidance also recognises discipline-specific methods and limitations.

This British framework is not binding on a Bangladeshi tribunal. Nor can it be applied mechanically without examining the laboratory’s data, model, population assumptions and validation procedures.

It does, however, demonstrate why a numerical result requires careful explanation.

A finding of limited statistical support is not equivalent to a finding that the voice belongs to someone else. Equally, limited support should not be presented as scientifically conclusive identification.

PW-45’s cross-examination introduces another question. The written record indicates that he could not identify the senior person said to have checked his findings and that the submitted report did not contain that reviewer’s signature.

That does not prove the examination was defective. It does identify a verification trail whose completeness matters when evaluating the weight to be attached to the expert opinion.

The same principle applies to the video examination discussed in PW-42’s testimony: finding no evidence of AI generation is one technical conclusion; identifying the original file, documenting its transfer and establishing the extent of the examination are additional tasks.

These questions have already been explored in important independent research. Taqbir Huda’s analysis for Tech Global Institute examines forensic methods, disclosure, defence limitations and digital-evidence standards in Bangladesh’s atrocity proceedings.

The present inquiry goes one stage further.

It asks how the technical results were used to support specific judicial findings, and whether comparable standards appear in other judgments made under the same legal framework.

The Law Was Already There

Some of the most consequential questions in this case arise not from international criticism, but from the tribunal’s own governing statute.

On 24 November 2024, Bangladesh amended the International Crimes (Tribunals) Act, 1973. The changes included provisions governing digital evidence, disclosure and the protection of accused persons.

Section 19(1B) authorised the tribunal, either on a party’s application or on its own initiative, to assess evidentiary relevance and admissibility, taking account of probative value and potential prejudice to a fair trial.

Section 19(1C) established when objections should ordinarily be raised. Section 19(1D) required the tribunal to give reasons for its evidentiary rulings.

Most significantly, Section 19(1F) provided for the exclusion of evidence obtained through violations of internationally recognised human rights where the violation substantially undermined reliability or admission would seriously damage the integrity of proceedings.

The provision resembles Article 69(7) of the Rome Statute of the International Criminal Court.

That similarity matters. It means the argument does not depend entirely on importing foreign principles into Bangladesh’s legal system. The domestic statute itself incorporated a comparable test.

But the test is conditional.

It would be legally incorrect to declare a recording inadmissible merely because it came from a surveillance agency, or because a leaked version circulated online.

The relevant human-rights violation must be established, followed by the statutory assessment of its consequences.

Equally, proof that a voice is authentic does not answer the separate question of whether the recording was obtained lawfully.

And a defence lawyer’s failure to make a timely objection does not erase the text of Section 19(1B), which expressly recognised the tribunal’s ability to consider admissibility on its own initiative.

The existence of that authority does not mean the court was required to initiate every conceivable inquiry or order every additional test requested by the defence.

The proper documentary examination concerns what the court considered, what it decided, and what reasons appear in the judgment and underlying evidentiary orders.

The tribunal was not bound by every technical rule of ordinary evidence law. That flexibility, however, is not the same as an exemption from the safeguards specifically written into its own statute.

From the Recorded Words to the Death Sentence

The recordings’ evidentiary significance cannot be understood without separating the tribunal’s findings under different charges.

The first charge involved the July statements and the killing of Abu Sayed. Under that charge, Hasina received imprisonment until natural death.

The death sentence was imposed under charges two and three, involving orders associated with the use of lethal force and the killings at Chankharpul and Ashulia on 5 August 2024.

This distinction matters for any challenge based on the disputed interpretation of the earlier phrase commonly translated as “I have already instructed.”

Even a significant ambiguity in the interpretation of that particular statement would not, on its own, invalidate every finding supporting the death sentence under the other charges.

The 18 July conversation is different. The references to lethal weapons, drones and helicopters are grave. They cannot responsibly be explained away as harmless political rhetoric.

But identifying the gravity of a statement is only the beginning of the criminal-law analysis.

The judgment must connect the relevant conduct to the legal basis of liability.

First, what order was given, and to whom? Second, how was that order transmitted or implemented? Third, which acts and omissions establish Hasina’s personal responsibility for the crimes charged?

The testimony of former police chief Al-Mamun, PW-36, is important to that inquiry. He described receiving information from the then home minister about instructions attributed to Hasina and relaying directions within the police hierarchy.

His evidence must be evaluated according to what he personally heard and did, and what he learned from others.

Those are not interchangeable sources of knowledge.

Nor is the prosecution’s case confined to his account. Other witnesses gave evidence about police orders and events on the ground. The tribunal also considered documentary, medical, video and other material.

This is why an evidentiary challenge cannot be reduced to removing one recording and declaring the entire prosecution case extinguished.

The tribunal’s statute also recognises superior responsibility. Section 4(3) is not limited to an express order to kill. It addresses forms of command or leadership responsibility, including specified failures to exercise control or prevent crimes in circumstances meeting the statutory conditions.

The decisive judicial question is whether the facts proved satisfy the elements of the particular mode of liability relied upon.

An evidential audit should therefore distinguish the proof that killings occurred, the proof of operational instructions and the proof connecting those crimes to the individual accused.

If a recording’s reliability or admissibility were successfully challenged, the consequences would depend on which findings required that recording and which remained supported independently.

That is a more demanding question than whether the prosecution possessed sufficient evidence to make a serious case.

It is also the question that makes the eventual judgment more, rather than less, accountable to the victims.

When One Tribunal’s Judgments Examine Another

A comparison of later tribunal judgments provides an unusual way of testing how digital evidence is understood within Bangladesh’s own judicial practice.

In its April 2026 Abu Sayed judgment, Tribunal-2 discussed the limits of a forensic voice examiner’s evidence. The court recognised that technical authentication could support an assessment of an audio recording without itself establishing an accused person’s involvement in the underlying crimes.

In February 2026, the Ashulia judgment subjected defence-produced digital materials to scrutiny over authorship, metadata, authenticity and chain of custody.

In June 2026, a Rampura judgment discussed a Sheikh Hasina–Fazle Noor Taposh conversation and described the examined voices as being quite similar to known voice samples.

None of these observations automatically contradicts the November 2025 judgment.

The exhibits, parties, experts, objections and purposes of the evidence differed. The fact that one tribunal questioned an electronic file’s provenance does not compel another tribunal to reject a different recording.

But judicial consistency is a legitimate subject of investigation.

If the provenance of a digital file matters when the defence produces it, the provenance of a digital file submitted by the prosecution is also relevant to determining its evidentiary weight.

If an expert’s voice identification does not itself establish criminal responsibility in one case, the distinction between identification and culpability remains analytically important in another.

And where an apparently related recording appears in two cases, a comparative audit can ask whether the same digital source, reference samples and forensic examination were used, and whether the resulting opinions were assigned comparable weight.

A different kind of question emerges near the end of Hasina’s November 2025 judgment.

In paragraphs 55–57 on printed page 444, the tribunal referred to interviews published in Indian media on 7 and 8 November 2025. One passage quoted Hasina as accepting ultimate leadership responsibility. Another characterised an interview as an admission of superior command responsibility.

These statements require careful treatment.

A political leader’s acceptance of general responsibility for events under her government is not necessarily an admission of every legal element of superior criminal responsibility.

Whether the particular words amounted to such an admission depends on their full context, the applicable law and the reasoning by which the tribunal interpreted them.

The chronology is also consequential. Published accounts place the conclusion of final arguments in October 2025, before the November interviews were published.

The issue is not whether judges are prohibited from noticing a later public statement. It is what evidentiary use was made of that statement, whether it contributed to findings of guilt or punishment, and whether the defence had an appropriate opportunity to address any new material relied upon.

The mere presence of the interviews in the judgment does not establish a procedural violation. The complete hearing record is needed to determine whether they were formally introduced, how they were considered, and whether the defence had an opportunity to respond.

But their appearance is a legitimate and specific subject for further judicial-document examination.

A Death Sentence and the Limits of Appeal

By November 2026, the judicial controversy had acquired a new dimension.

Hasina, living in India, announced an intention to return to Bangladesh and surrender. Lawyers acting on her behalf contacted the Bangladesh High Commission in New Delhi seeking travel and identity documents.

The government responded with sharply different public formulations. Some officials emphasised examination of the application’s procedural validity. Others insisted that extradition was the only acceptable route.

The dispute exposes the relationship between administrative powers, constitutional protections and the tribunal’s appeal provisions.

Section 21(3) of the International Crimes (Tribunals) Act prescribes a 30-day period for appeal and states that no appeal shall lie after its expiry.

The statutory language is restrictive. A defendant’s return from abroad does not automatically restart the clock.

There is, however, a difference between stating what a law provides and treating every possible question about its application as already conclusively decided.

The case of Abul Kalam Azad illustrates that difficulty.

Azad, sentenced to death in absentia in 2013 for crimes against humanity committed during the 1971 Liberation War, later returned and surrendered after an executive suspension of his sentence. He sought to challenge the judgment after a delay of many years.

The Appellate Division considered the question of whether his late appeal could be entertained. In September 2026, the court also cautioned the chief prosecutor against commenting on a matter pending before it.

The comparison does not create an automatic right for Hasina to a delayed appeal. Executive suspension of a sentence, judicial condonation of delay and acceptance of an appeal are different legal acts.

Recent reporting on Azad’s case is also not entirely consistent about its procedural outcome. A definitive comparison requires the latest certified Appellate Division order.

But the institutional principle is clear: the executive and prosecution may advance their interpretation of the law; the courts determine the legal consequences in cases properly placed before them.

The constitutional picture requires similar precision.

Article 47A of Bangladesh’s Constitution imposes exceptional restrictions affecting persons covered by laws under Article 47(3), including limits on specified constitutional rights and access to constitutional remedies before the Supreme Court.

That provision is severe. It should not be minimised.

But it is not accurately described simply by declaring that every fundamental or statutory protection disappears upon conviction.

The tribunal statute itself sets out rights relating to defence, counsel, evidence and appeal, subject to their terms. Constitutional restrictions, statutory rights and available judicial remedies must therefore be examined separately.

International standards offer an additional test rather than an automatic means of overturning the judgment. The UN Human Rights Committee’s General Comment No. 36 explains the especially serious consequences of violations of fair-trial guarantees in proceedings resulting in a death sentence.

That principle applies regardless of the identity or political history of the defendant.

The Difference Between Bringing Her Back and Bringing Her to Justice

The November 2026 dispute over Hasina’s passport adds a further layer.

The Bangladesh Passport Order, 1973 contains both discretionary and mandatory grounds on which a passport or travel document may be refused. It also provides procedures concerning refusal, revocation and the recording of reasons, subject to statutory exceptions.

A conviction or pending proceedings may provide grounds for restrictions. But the legal analysis depends on the particular provision, the applicant’s circumstances and any decision made by the competent authority.

A public statement that no lawful option exists should therefore be distinguished from a reasoned decision identifying the legal provisions that actually govern the request.

Hasina is not entitled to special treatment merely because she once held the office of prime minister.

Neither should the legal questions disappear because she is politically controversial.

Bangladesh may pursue extradition under its treaty with India. That process, however, involves Indian legal and executive decisions; a request from Dhaka does not by itself guarantee surrender.

A voluntary return would involve different practical requirements, including lawful departure from India, acceptable entry arrangements and appropriate custody procedures in Bangladesh.

None of these steps would automatically vacate the conviction, guarantee a new trial or create a fresh right of appeal.

What a government can responsibly consider is a lawful process that brings the person within the jurisdiction of the courts while preserving whatever rights and remedies remain available under law.

Such a process should also specify how any outstanding judicial questions will be placed before the competent court, without an executive promise to dictate the result.

If Hasina refuses a legally viable return arrangement, her declared intention to surrender would face a direct test.

If she returns, the credibility of Bangladesh’s institutions would face one of their own.

The two tests are not contradictory.

Nor should the country’s responsibility to investigate and punish atrocities be converted into a requirement to deprive an accused person of every meaningful procedural safeguard.

Who Watches the Watchers When the Recording Becomes Evidence?

The history of this case contains a striking institutional reversal.

A surveillance system developed during Hasina’s years in power helped produce material later used in proceedings against her.

That reversal makes the old question—Who watches the watchers?—more relevant, not less.

But the question now has several parts.

Who authorised the interception? Who preserved the original files? Who selected the recordings supplied to the tribunal? Who examined their authenticity? Who assessed the significance of the forensic findings? And who determined the relationship between the recorded words and the criminal responsibility ultimately attributed to the accused?

The answer cannot be that each institution’s assertion validates the next.

A properly functioning evidentiary system requires every important transition to be capable of examination: from interception to collection, from collection to forensic analysis, from analysis to evidentiary weight, and from evidentiary findings to criminal liability.

The existing record offers substantial evidence of atrocities, important forensic support for recordings and a detailed judgment assigning responsibility. It also identifies questions about provenance, disclosure, the communication of expert findings, the interpretation of statements and the application of the tribunal’s own safeguards.

Those questions warrant scrutiny. They do not, without further evidence, establish fabrication, deliberate judicial bias or the invalidity of the conviction.

Nor should they be dismissed on the ground that the accused is deeply unpopular or has already been condemned.

A death sentence is the most irreversible exercise of penal power. Its justification demands more than public certainty about a defendant’s political responsibility. It demands a legally sufficient explanation of individual criminal guilt, supported by admissible and properly evaluated evidence.

The families of those killed in July and August 2024 deserve that explanation no less than the accused deserves a legally fair process.

A judgment that withstands rigorous scrutiny serves victims better than one whose evidentiary foundations remain vulnerable to unresolved questions.

Bangladesh’s courts, lawmakers and public institutions do not have to choose between accountability for mass violence and the safeguards of justice.

They have to demonstrate that both can survive the same examination.

The question is not whether the state was entitled to investigate Sheikh Hasina. It plainly was. The question is whether every decisive step from a recorded conversation to a death sentence can withstand the legal scrutiny that the state’s own laws require.

A recording can preserve a voice. Only a fully reasoned judicial process can establish what that voice proves.

---

Reporting and editorial note

This is an independent documentary and legal analysis, not a judicial finding or a statement that any accused person has been exonerated. It distinguishes findings recorded in judgments, testimony, independently published investigations, statutory provisions and unresolved questions requiring further documentary verification.

The original trial judgment, selected subsequent judgments, Bangladesh’s legislation, international forensic guidance and independent reporting informed this analysis. Uncertified witness translations and media reporting are distinguished from authenticated judicial orders where their legal status affects the conclusion.

Bengal Gazette is an independent South Asia-focused publication operating through a UK-registered publishing organisation. Its organisational jurisdiction does not replace the legal requirements applicable to reporting, publication or distribution in other relevant jurisdictions.

Syndication and licensing

For editorial syndication, republication, documentary licensing or authorised use of Bengal Gazette investigations, contact:

Bengal Gazette — Editorial & Licensing

Email: editor@bengalgazette.asia

Telephone: +44 7352 771551

Postal correspondence: 82 Eve Road, West Ham, London E15 3DH, United Kingdom.

Original reporting, analysis and editorial materials are subject to applicable copyright and licensing terms.

The State’s Recording, the Court’s Reasoning, and the Unfinished Test of Justice

A Bengal Gazette Investigative Legal Analysis

By Abeer Uz Zaman

Bangladesh’s former prime minister was sentenced to death after a trial that relied on intercepted conversations, eyewitnesses, forensic examinations and evidence of killings during the 2024 uprising. The recordings have substantial corroboration. But the law asks questions beyond whether a voice is genuine: how evidence was obtained, what the science actually established, how individual criminal responsibility was inferred, and whether the defence had a meaningful opportunity to challenge that inference. A comparison of four tribunal judgments reveals why those distinctions matter.

---

The Objection That Came Too Late

On 21 October 2025, with Bangladesh’s International Crimes Tribunal approaching the end of a trial that would condemn a former prime minister to death, the lawyer appointed to defend Sheikh Hasina returned to the telephone recordings at the centre of the prosecution’s case.

The Criminal Investigation Department had examined the recordings. But Md Amir Hossain wanted an independent international examination. A tribunal judge asked why he had not raised the issue while questioning the forensic witnesses.

The lawyer acknowledged the omission.

“I could have mentioned it. I didn’t.”

The tribunal chairman declined to order another examination. Earlier in the exchange, a judge had pointed to examinations reported by the BBC and Al Jazeera.

It was a brief confrontation in a case overshadowed by the extraordinary violence of July and August 2024. Yet it exposed a question with consequences extending far beyond one defendant: when a lawyer fails to scrutinise an important piece of evidence adequately, what responsibility remains with the court?

The difficulty was not simply that the defence had raised its objection late. Under the tribunal’s governing law, evidentiary objections ordinarily had to be raised when evidence was submitted, with a limited exception for matters discovered later. But the same law empowered the court to consider admissibility on its own initiative and required reasons for evidentiary rulings.

Those provisions were already in force when the lawyer made his request.

On 17 November 2025, the tribunal convicted Hasina of crimes against humanity and sentenced her to imprisonment until natural death under the first charge and to death under the second and third charges. Former home minister Asaduzzaman Khan Kamal received the same two categories of punishment. Former police chief Chowdhury Abdullah Al-Mamun, who had become a prosecution witness, was sentenced to five years.

The judgments followed a national catastrophe. A United Nations fact-finding investigation estimated that as many as 1,400 people may have been killed between 15 July and 5 August 2024, most of them by security forces. Its findings described a systematic campaign of violent repression involving the former government and security apparatus.

The deaths, the injuries and the suffering of families are not speculative matters to be dismissed by disputing a particular recording. Nor does an allegation of procedural unfairness erase the prosecution’s substantial evidence.

But establishing that atrocities occurred and establishing the criminal responsibility of a particular defendant are different judicial tasks.

That distinction is indispensable when the punishment is death.

The State That Recorded Its Own Prime Minister

The telephone evidence has an institutional history that begins before the tribunal received a single recording.

Bangladesh’s National Telecommunication Monitoring Centre, or NTMC, was part of a surveillance architecture developed under successive governments, including Hasina’s own administration. Its capabilities, oversight and procurement history were examined in Bengal Gazette’s earlier investigation into the surveillance state.

That investigation asked a question familiar to democratic accountability: Who watches the watchers?

Hasina’s trial gave the question an unexpected consequence. Recordings associated with the state surveillance system that operated during her government later became evidence against her.

There is no inherent contradiction in that development. Evidence lawfully collected under one administration may properly be used to investigate wrongdoing by officials of that administration. Public authority must remain accountable even when power changes hands.

But the institutional reversal does not settle the legality of the original surveillance.

According to the testimony of special investigating officer Tanvir Hasan Zoha, identified as PW-53, investigators seized 69 audio clips and call-detail records relating to three telephone numbers from NTMC on 23 December 2024. Review began the following day. Selected conversations were later examined and introduced through the tribunal’s evidentiary process.

The distinction between collection and interception is essential. An investigator’s lawful seizure of an existing recording does not, by itself, establish the legal authority under which the original conversation was intercepted.

That does not mean the interception was unlawful. The telecommunications legislation operating in 2024 provided a statutory framework for authorised interception in specified circumstances involving national security and public order.

The unresolved documentary question is more precise: what authorisation applied to these particular calls, to which numbers, for what period, and under whose authority?

A second question arises from the selection of evidence.

Five conversations were identified in the investigator’s account as having been presented on three CDs. Selecting relevant material from a larger collection is ordinary prosecutorial practice. It becomes an evidentiary concern only if the selection removes material context, conceals relevant exculpatory information or prevents an adequate assessment of the recordings that were submitted.

The prosecution was not automatically obliged to play all 69 conversations in public. Some could contain irrelevant or sensitive communications involving third parties.

But the tribunal law contained a specific disclosure obligation. Section 12(2), introduced through the November 2024 amendment, required prosecutors to disclose, as soon as practicable, evidence in their possession that they believed might show innocence, mitigate guilt or affect the credibility of prosecution evidence.

The decisive issue is therefore not merely why five conversations were selected.

It is whether the available record establishes that the defence received the relevant context and material needed to test those five conversations.

Forensic authenticity begins with the file that an examiner receives. Evidentiary accountability begins earlier, with the history of how that file came into existence and reached the examiner.

What the Science Proved—and What It Did Not

The strongest argument against dismissing the recordings is that their authenticity was investigated.

The BBC’s 2025 investigation commissioned an independent examination by the audio-forensic organisation Earshot. That examination found no evidence of artificial synthesis or editing in the leaked recording it analysed.

The examination also suggested that the call may have been played through a loudspeaker and recorded by another device.

Those findings provide meaningful support for the authenticity of the material examined by Earshot.

They do not establish that the BBC’s leaked file was digitally identical to the files obtained from NTMC and examined for the tribunal.

That relationship needs to be demonstrated, not assumed.

An original interception recording, an exported file, a recording made from a loudspeaker and a later copy may contain the same conversation without being the same forensic exhibit.

Each has its own provenance. Each may have a different creation history, encoding, metadata and chain of custody. A test of one version cannot automatically authenticate every other version.

The distinction becomes especially important where a court treats the results of separate examinations as mutually reinforcing.

The Criminal Investigation Department’s work presents a related, but distinct, question.

The tribunal’s written judgment records the testimony of CID voice examiner Shahad Jubayer Lawrence, PW-45. In his examination of the Sheikh Hasina–Maksud Kamal conversation, the reported likelihood ratios for the compared voices were approximately 2.75 and 4.64.

Those figures have a specific scientific meaning.

A likelihood ratio compares how probable particular observations are under competing propositions. It is not the probability that the speaker is the alleged person. Nor is a ratio greater than one equivalent to identification beyond doubt.

The distinction is not a journalistic invention. The Forensic Science Regulator of England and Wales explains it in its guidance on interpretation and communication of forensic evidence.

Under that guidance’s illustrative scale, a likelihood ratio in the region of one to three may not materially distinguish between competing propositions, while a ratio between three and ten provides only slight support for one proposition over the other. The guidance also recognises discipline-specific methods and limitations.

This British framework is not binding on a Bangladeshi tribunal. Nor can it be applied mechanically without examining the laboratory’s data, model, population assumptions and validation procedures.

It does, however, demonstrate why a numerical result requires careful explanation.

A finding of limited statistical support is not equivalent to a finding that the voice belongs to someone else. Equally, limited support should not be presented as scientifically conclusive identification.

PW-45’s cross-examination introduces another question. The written record indicates that he could not identify the senior person said to have checked his findings and that the submitted report did not contain that reviewer’s signature.

That does not prove the examination was defective. It does identify a verification trail whose completeness matters when evaluating the weight to be attached to the expert opinion.

The same principle applies to the video examination discussed in PW-42’s testimony: finding no evidence of AI generation is one technical conclusion; identifying the original file, documenting its transfer and establishing the extent of the examination are additional tasks.

These questions have already been explored in important independent research. Taqbir Huda’s analysis for Tech Global Institute examines forensic methods, disclosure, defence limitations and digital-evidence standards in Bangladesh’s atrocity proceedings.

The present inquiry goes one stage further.

It asks how the technical results were used to support specific judicial findings, and whether comparable standards appear in other judgments made under the same legal framework.

The Law Was Already There

Some of the most consequential questions in this case arise not from international criticism, but from the tribunal’s own governing statute.

On 24 November 2024, Bangladesh amended the International Crimes (Tribunals) Act, 1973. The changes included provisions governing digital evidence, disclosure and the protection of accused persons.

Section 19(1B) authorised the tribunal, either on a party’s application or on its own initiative, to assess evidentiary relevance and admissibility, taking account of probative value and potential prejudice to a fair trial.

Section 19(1C) established when objections should ordinarily be raised. Section 19(1D) required the tribunal to give reasons for its evidentiary rulings.

Most significantly, Section 19(1F) provided for the exclusion of evidence obtained through violations of internationally recognised human rights where the violation substantially undermined reliability or admission would seriously damage the integrity of proceedings.

The provision resembles Article 69(7) of the Rome Statute of the International Criminal Court.

That similarity matters. It means the argument does not depend entirely on importing foreign principles into Bangladesh’s legal system. The domestic statute itself incorporated a comparable test.

But the test is conditional.

It would be legally incorrect to declare a recording inadmissible merely because it came from a surveillance agency, or because a leaked version circulated online.

The relevant human-rights violation must be established, followed by the statutory assessment of its consequences.

Equally, proof that a voice is authentic does not answer the separate question of whether the recording was obtained lawfully.

And a defence lawyer’s failure to make a timely objection does not erase the text of Section 19(1B), which expressly recognised the tribunal’s ability to consider admissibility on its own initiative.

The existence of that authority does not mean the court was required to initiate every conceivable inquiry or order every additional test requested by the defence.

The proper documentary examination concerns what the court considered, what it decided, and what reasons appear in the judgment and underlying evidentiary orders.

The tribunal was not bound by every technical rule of ordinary evidence law. That flexibility, however, is not the same as an exemption from the safeguards specifically written into its own statute.

From the Recorded Words to the Death Sentence

The recordings’ evidentiary significance cannot be understood without separating the tribunal’s findings under different charges.

The first charge involved the July statements and the killing of Abu Sayed. Under that charge, Hasina received imprisonment until natural death.

The death sentence was imposed under charges two and three, involving orders associated with the use of lethal force and the killings at Chankharpul and Ashulia on 5 August 2024.

This distinction matters for any challenge based on the disputed interpretation of the earlier phrase commonly translated as “I have already instructed.”

Even a significant ambiguity in the interpretation of that particular statement would not, on its own, invalidate every finding supporting the death sentence under the other charges.

The 18 July conversation is different. The references to lethal weapons, drones and helicopters are grave. They cannot responsibly be explained away as harmless political rhetoric.

But identifying the gravity of a statement is only the beginning of the criminal-law analysis.

The judgment must connect the relevant conduct to the legal basis of liability.

First, what order was given, and to whom? Second, how was that order transmitted or implemented? Third, which acts and omissions establish Hasina’s personal responsibility for the crimes charged?

The testimony of former police chief Al-Mamun, PW-36, is important to that inquiry. He described receiving information from the then home minister about instructions attributed to Hasina and relaying directions within the police hierarchy.

His evidence must be evaluated according to what he personally heard and did, and what he learned from others.

Those are not interchangeable sources of knowledge.

Nor is the prosecution’s case confined to his account. Other witnesses gave evidence about police orders and events on the ground. The tribunal also considered documentary, medical, video and other material.

This is why an evidentiary challenge cannot be reduced to removing one recording and declaring the entire prosecution case extinguished.

The tribunal’s statute also recognises superior responsibility. Section 4(3) is not limited to an express order to kill. It addresses forms of command or leadership responsibility, including specified failures to exercise control or prevent crimes in circumstances meeting the statutory conditions.

The decisive judicial question is whether the facts proved satisfy the elements of the particular mode of liability relied upon.

An evidential audit should therefore distinguish the proof that killings occurred, the proof of operational instructions and the proof connecting those crimes to the individual accused.

If a recording’s reliability or admissibility were successfully challenged, the consequences would depend on which findings required that recording and which remained supported independently.

That is a more demanding question than whether the prosecution possessed sufficient evidence to make a serious case.

It is also the question that makes the eventual judgment more, rather than less, accountable to the victims.

When One Tribunal’s Judgments Examine Another

A comparison of later tribunal judgments provides an unusual way of testing how digital evidence is understood within Bangladesh’s own judicial practice.

In its April 2026 Abu Sayed judgment, Tribunal-2 discussed the limits of a forensic voice examiner’s evidence. The court recognised that technical authentication could support an assessment of an audio recording without itself establishing an accused person’s involvement in the underlying crimes.

In February 2026, the Ashulia judgment subjected defence-produced digital materials to scrutiny over authorship, metadata, authenticity and chain of custody.

In June 2026, a Rampura judgment discussed a Sheikh Hasina–Fazle Noor Taposh conversation and described the examined voices as being quite similar to known voice samples.

None of these observations automatically contradicts the November 2025 judgment.

The exhibits, parties, experts, objections and purposes of the evidence differed. The fact that one tribunal questioned an electronic file’s provenance does not compel another tribunal to reject a different recording.

But judicial consistency is a legitimate subject of investigation.

If the provenance of a digital file matters when the defence produces it, the provenance of a digital file submitted by the prosecution is also relevant to determining its evidentiary weight.

If an expert’s voice identification does not itself establish criminal responsibility in one case, the distinction between identification and culpability remains analytically important in another.

And where an apparently related recording appears in two cases, a comparative audit can ask whether the same digital source, reference samples and forensic examination were used, and whether the resulting opinions were assigned comparable weight.

A different kind of question emerges near the end of Hasina’s November 2025 judgment.

In paragraphs 55–57 on printed page 444, the tribunal referred to interviews published in Indian media on 7 and 8 November 2025. One passage quoted Hasina as accepting ultimate leadership responsibility. Another characterised an interview as an admission of superior command responsibility.

These statements require careful treatment.

A political leader’s acceptance of general responsibility for events under her government is not necessarily an admission of every legal element of superior criminal responsibility.

Whether the particular words amounted to such an admission depends on their full context, the applicable law and the reasoning by which the tribunal interpreted them.

The chronology is also consequential. Published accounts place the conclusion of final arguments in October 2025, before the November interviews were published.

The issue is not whether judges are prohibited from noticing a later public statement. It is what evidentiary use was made of that statement, whether it contributed to findings of guilt or punishment, and whether the defence had an appropriate opportunity to address any new material relied upon.

The mere presence of the interviews in the judgment does not establish a procedural violation. The complete hearing record is needed to determine whether they were formally introduced, how they were considered, and whether the defence had an opportunity to respond.

But their appearance is a legitimate and specific subject for further judicial-document examination.

A Death Sentence and the Limits of Appeal

By November 2026, the judicial controversy had acquired a new dimension.

Hasina, living in India, announced an intention to return to Bangladesh and surrender. Lawyers acting on her behalf contacted the Bangladesh High Commission in New Delhi seeking travel and identity documents.

The government responded with sharply different public formulations. Some officials emphasised examination of the application’s procedural validity. Others insisted that extradition was the only acceptable route.

The dispute exposes the relationship between administrative powers, constitutional protections and the tribunal’s appeal provisions.

Section 21(3) of the International Crimes (Tribunals) Act prescribes a 30-day period for appeal and states that no appeal shall lie after its expiry.

The statutory language is restrictive. A defendant’s return from abroad does not automatically restart the clock.

There is, however, a difference between stating what a law provides and treating every possible question about its application as already conclusively decided.

The case of Abul Kalam Azad illustrates that difficulty.

Azad, sentenced to death in absentia in 2013 for crimes against humanity committed during the 1971 Liberation War, later returned and surrendered after an executive suspension of his sentence. He sought to challenge the judgment after a delay of many years.

The Appellate Division considered the question of whether his late appeal could be entertained. In September 2026, the court also cautioned the chief prosecutor against commenting on a matter pending before it.

The comparison does not create an automatic right for Hasina to a delayed appeal. Executive suspension of a sentence, judicial condonation of delay and acceptance of an appeal are different legal acts.

Recent reporting on Azad’s case is also not entirely consistent about its procedural outcome. A definitive comparison requires the latest certified Appellate Division order.

But the institutional principle is clear: the executive and prosecution may advance their interpretation of the law; the courts determine the legal consequences in cases properly placed before them.

The constitutional picture requires similar precision.

Article 47A of Bangladesh’s Constitution imposes exceptional restrictions affecting persons covered by laws under Article 47(3), including limits on specified constitutional rights and access to constitutional remedies before the Supreme Court.

That provision is severe. It should not be minimised.

But it is not accurately described simply by declaring that every fundamental or statutory protection disappears upon conviction.

The tribunal statute itself sets out rights relating to defence, counsel, evidence and appeal, subject to their terms. Constitutional restrictions, statutory rights and available judicial remedies must therefore be examined separately.

International standards offer an additional test rather than an automatic means of overturning the judgment. The UN Human Rights Committee’s General Comment No. 36 explains the especially serious consequences of violations of fair-trial guarantees in proceedings resulting in a death sentence.

That principle applies regardless of the identity or political history of the defendant.

The Difference Between Bringing Her Back and Bringing Her to Justice

The November 2026 dispute over Hasina’s passport adds a further layer.

The Bangladesh Passport Order, 1973 contains both discretionary and mandatory grounds on which a passport or travel document may be refused. It also provides procedures concerning refusal, revocation and the recording of reasons, subject to statutory exceptions.

A conviction or pending proceedings may provide grounds for restrictions. But the legal analysis depends on the particular provision, the applicant’s circumstances and any decision made by the competent authority.

A public statement that no lawful option exists should therefore be distinguished from a reasoned decision identifying the legal provisions that actually govern the request.

Hasina is not entitled to special treatment merely because she once held the office of prime minister.

Neither should the legal questions disappear because she is politically controversial.

Bangladesh may pursue extradition under its treaty with India. That process, however, involves Indian legal and executive decisions; a request from Dhaka does not by itself guarantee surrender.

A voluntary return would involve different practical requirements, including lawful departure from India, acceptable entry arrangements and appropriate custody procedures in Bangladesh.

None of these steps would automatically vacate the conviction, guarantee a new trial or create a fresh right of appeal.

What a government can responsibly consider is a lawful process that brings the person within the jurisdiction of the courts while preserving whatever rights and remedies remain available under law.

Such a process should also specify how any outstanding judicial questions will be placed before the competent court, without an executive promise to dictate the result.

If Hasina refuses a legally viable return arrangement, her declared intention to surrender would face a direct test.

If she returns, the credibility of Bangladesh’s institutions would face one of their own.

The two tests are not contradictory.

Nor should the country’s responsibility to investigate and punish atrocities be converted into a requirement to deprive an accused person of every meaningful procedural safeguard.

Who Watches the Watchers When the Recording Becomes Evidence?

The history of this case contains a striking institutional reversal.

A surveillance system developed during Hasina’s years in power helped produce material later used in proceedings against her.

That reversal makes the old question—Who watches the watchers?—more relevant, not less.

But the question now has several parts.

Who authorised the interception? Who preserved the original files? Who selected the recordings supplied to the tribunal? Who examined their authenticity? Who assessed the significance of the forensic findings? And who determined the relationship between the recorded words and the criminal responsibility ultimately attributed to the accused?

The answer cannot be that each institution’s assertion validates the next.

A properly functioning evidentiary system requires every important transition to be capable of examination: from interception to collection, from collection to forensic analysis, from analysis to evidentiary weight, and from evidentiary findings to criminal liability.

The existing record offers substantial evidence of atrocities, important forensic support for recordings and a detailed judgment assigning responsibility. It also identifies questions about provenance, disclosure, the communication of expert findings, the interpretation of statements and the application of the tribunal’s own safeguards.

Those questions warrant scrutiny. They do not, without further evidence, establish fabrication, deliberate judicial bias or the invalidity of the conviction.

Nor should they be dismissed on the ground that the accused is deeply unpopular or has already been condemned.

A death sentence is the most irreversible exercise of penal power. Its justification demands more than public certainty about a defendant’s political responsibility. It demands a legally sufficient explanation of individual criminal guilt, supported by admissible and properly evaluated evidence.

The families of those killed in July and August 2024 deserve that explanation no less than the accused deserves a legally fair process.

A judgment that withstands rigorous scrutiny serves victims better than one whose evidentiary foundations remain vulnerable to unresolved questions.

Bangladesh’s courts, lawmakers and public institutions do not have to choose between accountability for mass violence and the safeguards of justice.

They have to demonstrate that both can survive the same examination.

The question is not whether the state was entitled to investigate Sheikh Hasina. It plainly was. The question is whether every decisive step from a recorded conversation to a death sentence can withstand the legal scrutiny that the state’s own laws require.

A recording can preserve a voice. Only a fully reasoned judicial process can establish what that voice proves.

---

Reporting and editorial note

This is an independent documentary and legal analysis, not a judicial finding or a statement that any accused person has been exonerated. It distinguishes findings recorded in judgments, testimony, independently published investigations, statutory provisions and unresolved questions requiring further documentary verification.

The original trial judgment, selected subsequent judgments, Bangladesh’s legislation, international forensic guidance and independent reporting informed this analysis. Uncertified witness translations and media reporting are distinguished from authenticated judicial orders where their legal status affects the conclusion.

Bengal Gazette is an independent South Asia-focused publication operating through a UK-registered publishing organisation. Its organisational jurisdiction does not replace the legal requirements applicable to reporting, publication or distribution in other relevant jurisdictions.

Syndication and licensing

For editorial syndication, republication, documentary licensing or authorised use of Bengal Gazette investigations, contact:

Bengal Gazette — Editorial & Licensing

Email: editor@bengalgazette.asia

Telephone: +44 7352 771551

Postal correspondence: 82 Eve Road, West Ham, London E15 3DH, United Kingdom.

Original reporting, analysis and editorial materials are subject to applicable copyright and licensing terms.