Intercept evidence from foreign language communications: Reliability and minimum standards in the interests of justice

Introduction

Law enforcement authorities commonly use communication surveillance as an intrusive means of collecting information during their investigations to prosecute criminal activity. Lawfully intercepted communications between drug dealer and client, between bribery demander and supplier, and among members of criminal organisations are a particularly useful source of information for police investigations. Covertly intercepted communications may inform law enforcement authorities about the roles, activities, motivations and plans of people involved in criminal activity. Furthermore, communication surveillance is not only an effective means of covertly collecting information, but it also enables police officers to investigate from a distance without placing themselves in harm's way.

Selected parts of intercepted written communications from various instant messenger services (e.g., Threema, PlayStation, EncroChat, Sky ECC, ANOM, etc.) or from oral intercepted communications obtained from covertly placed listening devices are extensively used as evidence in every EU and Common Law country, including Australia, Canada, France, Germany, New Zealand, Israel and the US. When law enforcement officials and the courts cannot understand the source language of a communication, a suitably qualified interpreter or translator must undertake the task of rendering it into a language that can be understood. There is a terminological muddle regarding how researchers refer to interpreters and translators providing this service (Capus and Havelka, 2022: 1821). In this paper, we propose the term Intercept Interpreter/Translator (IIT). An IIT may be a permanent staff member of a law enforcement agency, or a person engaged under contract with a law enforcement agency.

In many countries, IITs contribute to criminal investigations by producing what we refer to asa translatum (translata in the plural form, cf. Reiss and Vermeer, 2014: 140), a generic term for all products of interlingual transfer (Griebel and Havelka, 2024: 1). We propose universal application of the term translatum to disambiguate references to IIT output. In this sense, the authors propose that IIT output consisting of a verbatim written document in the source language, derived from the intralingual transfer of the intercepted source language in oral form, may be referred to as a transcript, whereas a written document in the target language, derived from the direct interlingual transfer of the intercepted source language in oral form, is more appropriately termed a translatum. Both the transcript and the translatum are forms of translata.

The two primary functions of these translata are to serve operational and evidentiary purposes. In an operational context, verbal and written summaries may be produced to provide time-sensitive reporting of relevant information to operational decision-makers. Translated Intercept Records (TIR) are comprehensive written records in the target language which may be produced for intelligence purposes or for use in court as Translated Intercept Evidence (TIE).

As such, translata may vary in form from one procedural stage to another and be shared across jurisdictions. For example, translata may be transferred via international mutual legal assistance programmes or exchanged within Joint Investigation Teams of Eurojust and the European Union Agency of Criminal Justice Corporation. Joint investigations have been particularly important in recent years due to increased use of encrypted communications by criminal networks (Eurojust, 2023). The tasks that IITs perform for operational and evidentiary purposes fall within the field of criminal law, interpreting (the interlingual transfer from oral to oral and oral to written) and translation (the interlingual transfer from written text to written text) studies, and the broader field of forensic linguistics (Eades et al., 2023).

This paper addresses the questions of what minimal standards are required to realise a satisfactory level of reliability of translata produced by IIT and how courts may assess reliability. Generally, prosecutors rely on the assumption that TIR are accurate prior to being admitted as TIE, when in fact serious errors of translation contained in TIE may remain undetected throughout the trial process (Gilbert, 2014: 138). In contrast to monolingual intercepted communications, the reliability of TIE cannot be easily assessed by the trier of fact when the audio recording of the intercepted conversations is in a language other than that normally used in the court. Hence, law enforcement agencies need to be able to provide reliable TIE to the trier of fact, and the trier of fact needs to be able to assess the reliability of the TIE. We define reliability of TIE here as the extent to which the integrity of the primary evidence is preserved.

TIE reliability is explained in detail under ‘Reliability of TIE’. Courts frequently refer to notions of translation accuracy based on the premise that the translation from intercepted communications is a process of word-for-word transfer, including syntactic structure, from one language to another. When the audio recording of an intercepted communication is accompanied by a translatum, there is a commonly held misconception among judges and jury members that word-for-word translations are possible to ensure integrity of the evidence (Capus and Griebel, 2021: 92; Gilbert, 2014: 261; Gilbert and Heydon, 2021: 9). Based on current research findings, we consider that this idea of accuracy is not an adequate approach to assess the reliability of TIE noting that the application of context is essential in any translation process (Viaggio, 1992: 32). This is especially relevant when producing TIE to assist the trier of fact to understand the evidence contained in audio recordings of intercepted conversations. Translation approaches that remain close to form, or translate word-for-word, without adequate explanation by the translator in the form of notes where necessary, risk rendering the resultant translatum incomprehensible to the trier of fact. The IIT's dilemma is to produce TIE conveying meaning of the source language while also preserving the integrity of the intercepted evidence as required by the court. The IIT may produce a translatum that does not convey naturalness in the target language but may be considered by the courts to satisfy evidentiary requirements of accuracy. This is because the TIE appears to mirror the words and syntactic structures of the source language. The tendency of IIT to translate close to form, or more literally, results in the TIE making sense in some sections but not in others. For this reason, the authors use reliability as a core concept to address the quality of TIE by identifying and proposing minimum standards. The minimum standards proposed in this paper offer a basis from which a uniform approach can be applied to evaluate the reliability of the evidence, aligning judicial expectations with translation practices.

As a basis from which to propose minimum standards of producing reliable TIE, the authors first describe in detail how law enforcement authorities engage the IIT to assist with ongoing investigations and the collection of evidence. This is important to understand the complexity of the activity and to shed light on the extent to which evidence is likely to be impacted through the translation process (Capus and Griebel, 2021: 74; Capus and Havelka, 2022: 1821; Gilbert, 2014; Havelka, 2024). Second, the authors elaborate on the concept of reliability considering the specificity of TIE, positing that for TIE to be reliable in a broad sense, evidentiary as well as translational requirements of reliability must be met (Capus et al., 2024). From this basis, the paper proposes minimum standards that should be respected when producing and using TIE to mitigate prospects of interpretation/translation error. Reliability is also crucial when IITs undertake operational interpreting tasks during investigations, as operational effectiveness and the interests of law enforcement officials and suspect(s) are all affected. Erroneous intercept interpreting may result in ineffective decision-making, placing law enforcement agents at elevated risk when engaged in operational activity, such as conducting raids or arrests. Therefore, there is also a risk that the investigation becomes inefficient, and ineffective when relying on ambiguous or misinformation contained in translatum. Fundamental rights of wrongfully suspected persons may also be violated. While this paper focuses on reliability of TIE, we underline the necessity of ensuring reliability of any translata that IITs produce during operational phases of the investigation. While we propose minimum standards in this paper to improve operational and evidentiary processes, it is not our intention to prescribe how minimum standards should be implemented or operationalised in individual legal systems. The paper ends with some concluding remarks emphasising the need to further align judicial expectations with translation practices.

IIT—Operational support and evidence gathering

Investigations into suspected or actual criminal activity may depend on time-sensitive information interpreted from live intercepted communications. Police often rely on information from intercepted phone calls or from listening devices to make an arrest, initiate a raid or take action to prevent injury, death or property damage. Law enforcement authorities may also engage the IIT to perform tasks to meet intelligence collection requirements. For example, communication surveillance in the form of telephone interception may be required to collect information about the activities of a particular individual or group of interest suspected of being engaged in criminal activity. In such cases, the IIT will produce summaries relating to targeted criminal activity. IIT may also assist intelligence analysts with breaking codes, establishing patterns of speech, language and behaviour, and/or determining the organisational structure of a particular group.

Activities specific to the IIT role includes monitoring, selecting, translating, deciphering and transcribing the intercepted communication in close cooperation with police officers (Capus and Havelka, 2022: 1830; Drugan, 2020: 318; González Rodríguez, 2015). Hence, the tasks that IITs undertake go well beyond the general practice of interpreting or translating, culminating in a hybrid activity, an activity sui generis (Capus and Havelka, 2021: 191; 2022: 1822). IITs face significant challenges due to the nature of intercept interpreting, even when intercepted conversations are clearly heard and expressed in simple language. IITs are not a party to the communication event that they are intercepting but are observing and recording. This is an important point because consistency and context are essential elements of communication considering that the source language speaker has expressed the utterance in a manner that is of optimal relevance to the intended recipient (Gutt, 1998: 51). Therefore, as an external observer, the IIT assumes the contextual understanding shared between participants in the communication event and intervenes where necessary in the translation process based on the IIT's assumption of what is meant by what is said in the source utterance (Gilbert, 2014: 134). This setting is in stark contrast to the tasks undertaken by community or court interpreters/translators, who usually engage in, and facilitate the exchange of, communication between participating actors (González et al., 2012: 967 ss.). Interpreters and translators in community or court settings can ask questions, seek clarification and take body language of the source language speaker into account. In contrast, the IIT may be unaware of the context within which the intercepted conversation takes place. IITs may also be confronted with indistinct speech and extralinguistic sounds, such as background noise, that may make the source language difficult to hear or to understand. In such circumstances, the IIT does not have the opportunity to seek clarification, request repeats or to observe non-verbal cues.

The role of the IIT in support of law enforcement operations is not dissimilar to that of a military cryptologic linguist, where both aim to provide usable information derived from the interpretation/translation of intercepted communications in support of operational objectives (Gilbert, 2014: 7). The point of departure in terms of output is where IITs are tasked with producing translata for evidence purposes in the form of TIE. Cryptologic linguists provide translations primarily for operational or intelligence purposes. Another clear distinction in terms of training is that cryptologic linguists are highly trained in communications interception and operational interpreting with the requirement that they have exceptional comprehension and analytic skills. In contrast, IITs are usually interpreters and translators who have not received specific training in the skills required of communications interception and operational interpreting (Gilbert, 2014: 124). Research has established the need for IITs to be trained in the skill sets required to perform the tasks of operational interpreting and producing TIR/TIE, and the requirement for the introduction of standards has been identified (Capus and Bally, 2020; Capus and Havelka, 2022: 1832; Gilbert, 2014: 264; Taibi and Martin, 2012: 79). This deficiency in IIT training is concerning, considering that law enforcement authorities throughout the world increasingly rely upon support from IIT to combat trans-national criminal activity. Criminal activity includes threats to national security, such as terrorism, cyber-attacks, and the illicit-drug trade (Grabosky and Stohl, 2010: 71). It follows that appropriate IIT training should not only be recognised as necessary to uphold the principles of justice but also to effectively contribute to national and global security.

The nature of operational interpreting and translating requires IITs to take all available information into account to provide the decision-maker with time-sensitive information of relevance to the operation/investigation. This may be provided in the form of short verbal or written summaries of what was heard or written in intercepted communications. In such cases, using translation strategies that remain faithful to the source language are not likely to produce a report that is usable in time-sensitive operations. The report needs to be clear, concise, accurate, relevant and, most importantly, reported in an appropriate format to support effective decision-making. Therefore, the appropriate translation strategy applied for time-sensitive operational reporting will usually take the form of a meaning-based approach, having considered all contextual information available to the IIT at the time of making the report. This strategy is necessary to optimise the value of information provided to the decision-maker, considering that the significance of information is influenced by its nature and the context within which it is interpreted (Heuer, 1999: 42).

IITs may also be required to assist law enforcement officials with complex and detailed analysis of intercepted communications. Criminals may use codes and veiled speech to disguise the true nature of their activities (Nunn, 2010: 34). Not unlike the role of a cryptologic linguist in the military, IITs are likely to be called upon to apply their language and analytic skills to crack cyphers and to assist with making sense of intercepted communications containing veiled speech or coded language (Gilbert, 2014: 255). The interpretation of intercepted text messages containing emojis, abbreviations in a mix of languages, and acronyms also presents challenges to IITs (Loiacono, 2022: 92) The methodology that the IIT applies when performing analytic tasks of this nature is dynamic. Translation approaches that remain true to the form of the source language and that apply a more dynamic approach conveying sense in the target language, are both necessary to determine the underlying meaning of coded language (Gilbert, 2014: 231).

Considering the distinction between IIT roles of operational interpreting and that of producing translata for evidentiary purposes, it is essential to identify the variance in approaches to the translation process for each purpose. When producing TIE, the IIT is required to produce a document in the language of the court that mirrors what the IIT perceived in the intercepted communication. In contrast to translata for operational purposes, TIE is neither perishable nor time-sensitive. Instead, courts generally require the TIE to be complete and not fused with intelligence or background contextual information. For evidentiary purposes, the courts require that TIE represents an accurate transfer of communication from the source into the target language (Gilbert, 2014: 223). This approach towards finding equivalence during the interlingual transfer process aligns with what may be referred to as a formal or more literal approach which seeks to retain, as closely as possible, the linguistic elements of the source text (Nida, 1964: 2). However, this approach has significant limitations when framed within the context of TIE, as Taibi and Martin (2012) show based on the case of Taysir Alony, the Al-Jazeera reporter who was imprisoned because of alleged collaboration with a terrorist organisation. Analysing several examples, Taibi and Martin demonstrate the inappropriateness of excessive literal translation (2012: 81). Literal translations that fail to take pragmatic and cultural aspects into account should be accompanied by translator's notes.

Misinformed notions of accuracy held by the courts create a serious dilemma for the IIT. Courts expect TIE to be narrowly translated and not to reflect the IIT's analytical and contextual perspective that is usually applied when supporting law enforcement operations. Translation studies have long established that interlingual transfer always includes interpretation (Nida, 1964: 156; Viaggio, 1992: 31). Additionally, it is generally accepted that no two languages are the same in terms of lexicon, semantics or syntactic structure; therefore, there can never be a perfectly accurate translation. While formal equivalents may not exist between the source language and the target language, an equivalent effect may be achievable (Nida, 1964: 156). A dynamic approach through translator intervention when producing TIE is likely necessary to convey sense.

Research involving linguistic analysis of TIE has demonstrated that IITs, like interpreters and translators in general, do not act as mere conduits (Angelelli, 2004; Gilbert, 2014; Laster and Taylor, 1995: 10). When the source language is translated word for word, the resultant translation is usually nonsensical in the target language. A translation approach that remains closely aligned to the source language structure and content without conveying sense can produce ambiguity and distortion in the target language. An example of the difference between a transcript of the source language (Vietnamese), a word-for-word translation into the target language (English), the corresponding TIE used in court (English) and a proposed correct translation (English), provided in the example below. The example is of a lawfully intercepted conversation using a listening device placed in a kitchen. Two people were present in the room, and one was dividing compressed heroin with a knife. Each time the person cut the block of heroin, some of it would crumble away. The person dividing the heroin became frustrated, and the proposed correct translation below reflects this. However, by remaining too close to the form of the source language (translating too literally), the IIT failed to convey sense, which resulted in significant confusion in the court room.

The following is an example of TIE containing idiomatic expression (Gilbert, 2014: 94):

OPEN IN VIEWER

Source language (Vietnamese)	Chia-là-mất,-chết.
Word-for-word translation	Divide-is-lose,-dead.
TIE	Lose it, God oh God, is it dead?
Proposed correct translation	Dividing (it) means losing (some), damn it!

Providing a word-for-word translation (line 2) of the source language transcript (line 1) demonstrates what one may expect of the IIT acting purely as a conduit. Using the example above, we can see that the translator has intervened in the translation process to render the language appearing in the TIE (line 3). However, in this example, the TIE is considered unreliable as is evidenced by the proposed correct translation (line 4). This example demonstrates how a narrow approach to translation can disadvantage both the prosecution and the defence, as the IIT has remained too close to the form of the source language, translating it in a rather awkward way. The literal meaning of the word ‘chết’ in Vietnamese is ‘dead’, as it appears in the TIE. In the context of the source utterance, the word ‘chết’ is an idiomatic term expressing the emotion of frustration as seen in the proposed correct translation where the word ‘chết’ is translated as ‘damn it!’ It became evident during the trial that even the prosecution was confused about who or what was dead.

During questioning, the prosecution asked the defendant, giving evidence in their defence, what they meant when they said in Vietnamese, ‘Lose it, God oh God, is it dead?’ The prosecutor rephrased the question and asked, ‘Who or what was dead?’ The Vietnamese court interpreter interpreted the question posed by the prosecutor with no scope to convey the idiomatic meaning of ‘chết’ as it appeared in the original utterance. The defendant was confused, and the prosecutor became frustrated. The court interpreter knew what the problem was but remained impartial for ethical reasons. The witness was unable to respond in a way that appeased the prosecutor because the idiomatic meaning was lost in the TIE, and therefore lost in the court interpreting process. Confusion continued in the courtroom until the issue was resolved with the assistance of another court interpreter present (Gilbert and Heydon, 2021: 7). The above example demonstrates the importance of IITs conveying sense wherever possible and providing translator's comments where necessary.

Previous research has shown that the IIT shifts between producing a translation remaining as close to the form of the original language as possible (often described as a literal approach), and a more flexible and creative approach that focuses on conveying sense in the target language (a free approach). The difference between these approaches is that the approach remaining close to the form of the target language often requires the intended readership to have some understanding of the linguistic and cultural context of the source language, in contrast to the dynamic or creative approach, which aims to produce a seamless translation that is linguistically and culturally adapted to the target readership. It is often necessary for a formal approach to be compromised to convey sense in the target language, in which case translators will need to provide notes or comments for the translation to make sense (Nida, 1964: 159). This is particularly relevant when the culture associated with the source language is vastly different from that of the target language (Nida, 1964: 161). The fluctuation between formal and dynamic translation approaches often results in incoherent sections of TIE. This is particularly prevalent when IITs translate word-for-word because they are in doubt as to what the intercepted words mean when context is unclear or unknown. Incoherent shifts in language style appearing in TIE make it difficult for the trier of fact to understand what is alleged to have been said and what was meant in the source language. Mistranslations contained in TIE may be attributed to IITs not having access to contextual information of relevance, resulting in word-for-word translations becoming more frequent. Frequent shifts between remaining true to the form of the source language and applying dynamic and creative approaches to convey meaning in the target language often result in awkward and disjointed language appearing in TIE. Such TIE may be challenged by the defence citing claims of unreliability due to the TIE being inaccurate or incorrect.

Reliability of TIE

TIE helps the trier of fact understand the source language (a language other than that of the court) contained in audio recordings and text obtained from telephone intercepts, covert listening devices and other sources (González et al., 2012: 965 ss). Intercepted audio recordings are generally referred to as primary evidence, and TIE is referred to as secondary evidence produced to assist the trier of fact (Fishman, 2006: 487). How the IIT approaches the task of producing TIE is a significant factor that will influence the trier of fact when assessing weight given to TIE. Research has shown that even monolingual transcripts of poor-quality audio recordings risk priming the perception of the trier of fact (Fraser et al., 2011; Fraser and Stevenson, 2014). In this context, priming can be described as the effect of a transcript influencing the perception of the trier of fact to align with that of what the transcript purports to say. In contrast to monolingual transcripts, TIE poses even a greater risk that the trier of fact will readily accept the secondary evidence presented in a language they understand, as they are unable to understand the primary evidence in the form of an audio recording in a language other than that of the court. Hence, two main shifts affect reliability when producing TIE: the transference from one language to another, and the shift from oral to written form. Because of these shifts, many factors contained in the primary evidence are inevitably lost, which would otherwise—in a monolingual situation—be perceived and understood by the trier of fact had the primary evidence been recorded in a language of the court (for a detailed analysis of this multi-layered translational process, see Havelka, 2024). Also, considering the large volume of information collected, the IIT will necessarily filter out non-relevant information and produce summaries so that the information reported is more suitable for operational decision-making.

Police officers, prosecutors, defence lawyers, judges and jury members depend on the translatum provided by IITs, which underlines the importance that translatum reliability criteria are established, mutually understood and implemented by all parties (González et al., 2012: 968 ss). To identify reliability criteria in a holistic context is complex but is essential in the pursuit of a fair trial. The particularity of TIE is derived from a hybrid nature of the translational activity, which is inextricably linked to criminal law procedure. For this reason, reliability criteria of TIE need to consider evidentiary requirements and translational methodologies used to produce TIE.

Due to the heterogenous nature of national evidentiary rules, the process of producing TIE varies across jurisdictions, not only between Civil Law and Common Law countries, but also between countries using the same legal system. To illustrate this variance, a few examples are briefly explored. For example, in the United Kingdom, intercept material stemming from communication surveillance cannot be adduced as evidence (cf. Regulation of Investigatory Powers Act 2000). There are, however, four exceptions: communication surveillance (including interception of phone calls), recordings from the use of listening devices, interceptions derived from other countries and intercepted phone calls made from prisons.

While in both Common Law and Civil Law countries the sounds and tones recorded on an audio file are considered by the court as primary evidence and the TIE as secondary evidence, the rules of evidence differ regarding whether audio records must be played aloud in court so the defendant has an opportunity to hear the primary evidence against them. Rules also differ considerably between jurisdictions regarding the point in time when defendants must be provided the opportunity to learn about the existence and content of the TIE.

Countries’ legal requirements may also vary in relation to the production of secondary evidence. In many countries, such as Australia, Austria, Germany and Switzerland, intercepted communications are directly transferred into the written target language (Capus and Havelka, 2021: 182s.). In contrast, the US has established a two-step process as the preferred method in accordance with guidelines published by the U.S. National Association for Judiciary Interpreters and Translators (NAJIT, 2019: 1). According to NAJIT guidelines, the intercepted oral communications are first transcribed into the written source language in accordance with prescribed conventions of transcription. The second step requires the transcription in the source language to be translated into the target language, usually the language of the court. In some countries, such as the US, the IIT may be called as a witness to establish the methods they used to produce the translatum (Fishman, 2006).

Evidence rules are notably different regarding the (in)admissibility of TIE. In Australia, for example, even if there exists the possibility or likelihood that evidence has been falsified, it may still be found to be relevant, and therefore admissible (Heydon, 2015: 105). In jury trials, judges consider the weight of relevant evidence within the context of the case. Relevant evidence under Australia's Uniform Evidence Law is defined as ‘evidence that, if it were accepted, could rationally affect (directly or indirectly) the assessment of the probability of the existence of a fact in issue in the proceeding’ (Evidence Act 1995 [Cth]—Section 55). Therefore, the threshold for admissibility of evidence is low (Heydon, 2015: 103). The rationale is that evidence of the slightest probative value should not be withheld from the jury unless the probative value of the evidence is outweighed by the danger of unfair prejudice to the defendant.

Notwithstanding such differences in relation to the reliability of TIE, the legal threshold required can be directly deduced from the internationally recognised principle of a fair trial. This principle has materialised into the fundamental right of a fair trial in the European Convention on Human Rights (Article 6), the American Convention on Human Rights (Article 8), the African Charter on Human and Peoples’ Rights and the accompanying Proclamation of the African Commission on Human and Peoples’ Rights of the Principles and Guidelines on the Right to a Fair Trial and Legal Assistance in Africa (African Union, 1992) and the International Covenant on Civil and Political Rights (Article 14; Momsen and Willumat, 2024). The principle underpins the criteria that must be met for reliable TIE. This is regardless of jurisdiction-specific evidentiary rules aimed at realising procedural fairness (cf. for the same approach taken in the context of digital evidence: Stoykova, 2023). Procedural fairness is concerned with fairness of the procedure leading to a decision, and not to the fairness of the decision itself (ALRC, 2013). Procedural fairness serves the interests of justice, balancing the interests of the defendant against those of the community. Therefore establishing minimum standards for intercept interpreting protects both the State and the individual based on the premise that the way in which evidence is obtained, and in this case translated, must be fair.

By intersecting fundamental legal requirements that underpin the principle of fair trial with translational requirements, the authors identify the following two reliability criteria for TIE: impartiality and verifiability (Capus et al., 2024).

Minimum standards

This paper proposes minimum standards to enhance the quality of IIT contributions to operational requirements and TIE reliability. The authors emphasise that ensuring reliability of TIR before it is presented as TIE is a critical factor that justifies the establishment of minimum standards. In the following presentation, three categories of minimum standards are proposed: general, translatory procedural and (pre-)trial minimum standards.