GIL v Public Prosecutor [2025] SGCA 21

Case Details

• Citation: [2025] SGCA 21
• Court: Court of Appeal of the Republic of Singapore
• Decision Date: 06 May 2025
• Coram: Tay Yong Kwang JCA; Belinda Ang Saw Ean JCA; Debbie Ong Siew Ling JAD
• Case Number: Criminal Motion No 42 of 2024
• Hearing Date(s): 7 March 2025
• Claimants / Plaintiffs: GIL (Applicant)
• Respondent / Defendant: Public Prosecutor
• Counsel for Claimants: Tan Jun Yin and Tanaya Kinjavdekar (Trident Law Corporation)
• Counsel for Respondent: Wuan Kin Lek Nicholas and Teo Siu Ming (Attorney-General’s Chambers)
• Practice Areas: Criminal Procedure and Sentencing; Evidence; Statutory Interpretation

Summary

The Court of Appeal in [2025] SGCA 21 addressed a critical intersection between criminal procedure and the law of evidence, specifically regarding the scope of statutory presumptions in the digital age. The applicant, GIL, sought permission under s 397(1) of the Criminal Procedure Code 2010 to refer a question of law of public interest to the Court of Appeal. This followed the dismissal of his Magistrate’s Appeal in [2024] SGHC 287, where he had unsuccessfully challenged a conviction for outrage of modesty under s 354(2) of the Penal Code. The crux of the application concerned the proper interpretation of s 116A(1) of the Evidence Act 1893, a provision designed to facilitate the admission of electronic records.

The central doctrinal contribution of this judgment lies in its clarification of the distinction between the "accurate communication" of an electronic record and the "substantive reliability" or "veracity" of the data contained therein. The applicant contended that s 116A(1) of the Evidence Act 1893 required the court to presume not only that a device was in working order but also that the data it generated—in this case, sleep-tracking data from a smartwatch—was an accurate and reliable reflection of reality. The Court of Appeal rejected this expansive reading, affirming that the presumption is a modern restatement of the common law maxim praesumuntur omnia rite esse acta. This maxim presumes that mechanical instruments are in order when used, but it does not extend to the truth of the substantive information produced by the device's internal algorithms or processes.

The appellate result was a dismissal of the criminal motion. The Court found no conflict of judicial authority between the High Court's decision in the applicant's case and prior authorities such as Super Group Ltd v Mysore Nagaraja Kartik [2019] 4 SLR 692 and Telemedia Pacific Group Ltd v Credit Agricole (Suisse) SA [2015] 1 SLR 338. The Court emphasized that the legislative intent behind the 2012 amendments to the Evidence Act 1893 was to streamline the admission of computer output, not to create a shortcut for proving the truth of complex data points generated by sophisticated software.

This decision has broader significance for practitioners dealing with Internet of Things (IoT) devices and wearable technology in litigation. It establishes that while the "output" of a device may be easily admitted into evidence via the s 116A(1) presumption, the "weight" and "reliability" of that output remain subject to traditional evidentiary challenges. In cases involving complex data generation, such as biometric tracking or algorithmic inferences, the court will not presume the data is "correct" simply because the device was functioning. This preserves the necessity for expert evidence to explain the underlying mechanics of how such data is derived and its inherent limitations.

Timeline of Events

1. 27 February 2021: The applicant’s daughter (the “Daughter”) and the victim (then 12 years of age) have a sleepover in the Daughter’s bedroom.
2. 28 February 2021: The alleged offence of outrage of modesty occurs in the early hours of the morning. The victim later identifies the time as approximately 1:57 am.
3. Trial (District Court): The applicant is tried and convicted of one charge of outrage of modesty under s 354(2) of the Penal Code. The District Court relies on the victim's testimony and rejects the applicant's defense.
4. 6 November 2024: The High Court (Magistrate’s Appeal) dismisses the applicant’s appeal against conviction. The Judge publishes written grounds in [2024] SGHC 287, addressing the relevance of the applicant's smartwatch data.
5. Post-Appeal: The applicant files Criminal Motion No 42 of 2024 (CM 42) seeking permission to refer a question of law to the Court of Appeal under s 397(1) of the Criminal Procedure Code 2010.
6. 7 March 2025: The Court of Appeal hears the substantive arguments for CM 42.
7. 7 March 2025: The Court of Appeal dismisses CM 42 immediately following the hearing.
8. 06 May 2025: The Court of Appeal delivers its detailed grounds of decision for the dismissal of the criminal motion.

What Were the Facts of This Case?

The factual matrix centers on an incident that occurred during a sleepover on 27 February 2021. The applicant’s daughter and the victim, a 12-year-old girl, were in the Daughter’s bedroom. According to the evidence accepted by the lower courts, the two girls were sitting on the top bunk of a bed, positioned under a duvet. They were using their mobile phones to engage with an application called "Combyne," which allows users to design outfits. The applicant was also present in the bedroom at the time.

The prosecution's case was that while the girls were occupied with the app, the applicant slipped his hand under the duvet and touched the victim’s thigh and vaginal area. The victim testified that this occurred at approximately 1:57 am on 28 February 2021. The applicant denied the allegations entirely, maintaining that no such contact took place and that he was asleep during the period in question.

A central piece of evidence introduced by the defense was data extracted from the applicant's "Amazfit GTR Smartwatch" (the “Watch”). The "Watch Data" became a focal point of the legal dispute. The applicant argued that the Watch Data showed he was in a state of "deep sleep" or "light sleep" at 1:57 am, which directly contradicted the victim’s account of the timing of the offence. The applicant’s position was that if he was asleep, he could not have committed the act alleged by the victim at the time she specified.

In the District Court, the parties had agreed to the admission of the Watch Data into evidence. However, the District Court found that the victim was a credible witness and that the discrepancy in timing did not undermine the core of her allegation. The High Court, in [2024] SGHC 287, upheld the conviction. The High Court Judge noted that the Watch Data did not conclusively prove the applicant was asleep, as the reliability of the watch's sleep-tracking algorithm was not established. Furthermore, the Judge held that even if the timing was slightly off, it did not mean the incident did not occur at all.

The procedural history then moved to the Court of Appeal via CM 42. The applicant sought to frame the High Court's treatment of the Watch Data as a misinterpretation of the statutory presumptions in the Evidence Act 1893. Specifically, he argued that the Judge had failed to apply the presumption under s 116A(1), which he claimed should have forced the court to accept the Watch Data as an accurate and reliable record of his sleep state. This set the stage for a deep dive into the mechanics of electronic evidence and the limits of what a court must "presume" when faced with digital output.

What Were the Key Legal Issues?

The application for a criminal reference required the Court of Appeal to determine whether the threshold requirements of s 397 of the Criminal Procedure Code 2010 were met. The primary legal issues were framed as follows:

• The Interpretation of s 116A(1) of the Evidence Act 1893: Does the presumption that an electronic record was "accurately produced or communicated" mean that the court must presume the substantive content of that record is true and reliable? Or does it merely presume that the device was in working order and accurately reflected what it was designed to output?
• The Existence of a Conflict of Judicial Authority: Was there a genuine conflict between the High Court's decision in the applicant's case and the decisions in Super Group Ltd v Mysore Nagaraja Kartik [2019] 4 SLR 692 and Telemedia Pacific Group Ltd v Credit Agricole (Suisse) SA [2015] 1 SLR 338? This was a prerequisite for the question to be deemed one of "public interest" under s 397(6)(a) of the Criminal Procedure Code 2010.
• The "Outcome" Requirement: Even if a question of law existed, did the High Court Judge’s determination of that question actually affect the outcome of the case? If the conviction would have stood regardless of the interpretation of s 116A(1), the reference would be academic.

These issues are critical because they define the boundary between the admissibility of digital evidence (which s 116A facilitates) and the weight or probative value of that evidence (which remains a matter for the trier of fact). The applicant’s attempt to merge these two concepts represented a significant challenge to established evidentiary principles.

How Did the Court Analyse the Issues?

The Court of Appeal’s analysis began with the statutory text of s 116A(1) of the Evidence Act 1893. The provision states that it shall be presumed, unless the contrary is proved, that the electronic record in question was "accurately produced or communicated" by the relevant device or process. The Court noted that the applicant’s proposed "Question" for reference was:

"Does section 116A(1) of the EA presume only that an electronic record in question is an accurate reflection of what was produced or communicated by the electronic device, or that the electronic record is accurate and reliable?" (at [12])

The Court conducted a thorough review of the legislative history of the 2012 amendments to the Evidence Act 1893. It noted that prior to these amendments, the admission of computer output was governed by a "cumbersome process" under the old sections 35 and 36, which required certificates of authenticity. The 2012 amendments replaced this with a framework that focused on the "working order" of the machine. The Court cited the speech of the Minister for Law, K Shanmugam, during the second reading of the Evidence (Amendment) Bill on 14 February 2012:

"The focus is thus on facilitating the admission of evidence produced by machines, specifically, by presuming that the machines which produced the relevant electronic records were in working order and thus accurately communicated the material." (at [19])

The Court concluded that s 116A(1) is a restatement of the common law maxim praesumuntur omnia rite esse acta. This maxim presumes that mechanical instruments were in order when they were used. However, the Court drew a sharp distinction between the machine's operation and the truth of its output. As the Court explained at [18]:

"It is clear from this that the presumption applies to the accurate communication of the electronic record, not the veracity or truth of the data contained in that record."

To illustrate this, the Court used the example of a thermometer. If a thermometer is in working order, s 116A(1) presumes it accurately "communicates" a reading of 38 degrees Celsius. It does not, however, presume that the person being measured actually has a fever of 38 degrees if, for instance, the thermometer was placed near a heat source or used incorrectly. The "accuracy" of the communication is distinct from the "reliability" of the underlying fact the data purports to represent.

The Court then addressed the alleged conflict with Telemedia and Super Group. In Telemedia, the High Court had held that the presumption in s 116A(1) relates to the "accurate communication" of the record. In Super Group, the court had similarly noted that the presumption facilitates the admission of computer output by presuming the device was in working order. The Court of Appeal found that these cases were entirely consistent with the High Court Judge’s approach in the present case. There was no judicial authority suggesting that s 116A(1) creates a presumption of substantive truth or reliability of the data content.

Furthermore, the Court delved into the complexities of modern technology, referencing Katherine B Forrest’s work on Artificial Intelligence. The Court noted that modern devices often use complex algorithms to "infer" data (such as sleep states) from raw sensor inputs (such as movement or heart rate). These inferences are not simple mechanical outputs like a photocopy or a printed log. At [40], the Court observed:

"The rationale for the presumption does not extend to accepting that just because evidence is produced by machines which were in order and used properly, the substantive data or content contained in the electronic record is true and reliable."

The Court held that the applicant was essentially asking the court to presume that the Watch’s internal algorithm for determining "sleep" was 100% accurate. This would be an impermissible leap. The presumption in s 116A(1) only goes so far as to say the Watch accurately recorded and communicated whatever data it was programmed to produce. Whether that data (the "sleep" label) was a reliable reflection of the applicant's actual physiological state was a separate question of weight, requiring expert evidence which the applicant had failed to provide.

What Was the Outcome?

The Court of Appeal dismissed the applicant’s motion for permission to refer the question of law. The Court found that the requirements of s 397(1) of the Criminal Procedure Code 2010 were not met because there was no "question of law of public interest." Specifically, there was no conflict of judicial authority, and the interpretation of s 116A(1) of the Evidence Act 1893 was already settled and consistent across the High Court and the Court of Appeal's own analysis of legislative intent.

The operative paragraph of the judgment states:

"For the reasons set out above, we dismissed the applicant’s application for permission to refer the Question to the Court of Appeal." (at [45])

Regarding the specific disposition for the parties:

• The Applicant (GIL): The application was dismissed in its entirety. His conviction under s 354(2) of the Penal Code and the sentence imposed by the lower courts remained undisturbed.
• The Respondent (Public Prosecutor): The Respondent successfully resisted the motion.

Costs: The Court followed the usual practice in criminal matters and made the following order:

"We made no order as to costs." (at [47])

The Court also noted that even if the question had been answered in the applicant's favor, it would not have affected the outcome of the case. The High Court Judge had already considered the Watch Data and found that, even if admitted and presumed to be an accurate communication, it did not create a reasonable doubt regarding the applicant's guilt. The victim's testimony was found to be sufficiently robust to overcome the discrepancy in timing suggested by the Watch Data. Thus, the reference would have been academic in any event.

Why Does This Case Matter?

This case is a landmark clarification of the limits of digital evidence in Singapore. As society becomes increasingly reliant on wearable technology, smart home devices, and algorithmic data, practitioners must understand that the Evidence Act 1893 does not provide a "magic wand" to prove the truth of digital assertions. The ratio of the case—that s 116A(1) presumes accurate communication but not substantive veracity—places the burden of proving reliability squarely on the party seeking to rely on the data.

From a doctrinal perspective, the judgment reinforces the traditional distinction between admissibility and weight. Section 116A(1) is an admissibility-facilitating provision. It removes the need for a technician to testify every time a computer printout is used, provided the machine is in working order. However, once the evidence is "in," the court must still evaluate its probative value. In the context of IoT devices, where "data" is often the result of proprietary, opaque algorithms (the "black box" problem), the Court of Appeal has signaled that it will not allow statutory presumptions to bypass the need for rigorous proof of reliability.

For criminal practitioners, the case serves as a warning against over-reliance on "digital alibis" without supporting expert testimony. The applicant’s failure to provide an expert to explain how the Amazfit GTR determines "sleep" was fatal to his defense. The court will not take judicial notice of the accuracy of consumer-grade sleep trackers. This underscores the necessity for defense counsel to engage forensic experts who can explain the error rates, sensor limitations, and algorithmic assumptions of the devices their clients rely upon.

Furthermore, the case clarifies the "conflict of authority" requirement for criminal references. The Court of Appeal made it clear that a mere difference in emphasis or a "novel" application of a settled principle does not constitute a conflict. This maintains the high threshold for references to the Court of Appeal, ensuring that the criminal reference process is not used as a "second appeal" on the facts or on settled law.

Finally, the judgment aligns Singapore’s approach with other common law jurisdictions, such as Australia. By noting the similarities between s 116A(1) and sections 146 and 147 of the Australian Commonwealth Evidence Act 1995, the Court has integrated Singapore’s digital evidence framework into a broader international context of "working order" presumptions. This provides a useful comparative baseline for future litigation involving cross-border digital evidence.

Practice Pointers

• Distinguish Admissibility from Weight: When relying on s 116A(1) of the Evidence Act 1893, remember that the presumption only gets the evidence through the door. It does not prove that the information contained in the record is true.
• Engage Experts for Algorithmic Data: If your case relies on data that is "inferred" by a device (e.g., sleep states, steps taken, heart rate variability), you must provide expert evidence to explain the reliability of the underlying algorithm. The court will not presume the algorithm is accurate.
• Challenge the "Working Order": Conversely, if opposing an electronic record, focus on rebutting the presumption that the device was in "working order" or that the communication was "accurate." If the device was damaged or the software was glitching, the s 116A(1) presumption can be rebutted.
• Threshold for Criminal References: Ensure that any question of law proposed for reference under s 397 of the Criminal Procedure Code 2010 is backed by a genuine conflict of judicial authority. Disagreement with a judge's application of settled law to facts is insufficient.
• Verify Data Sources: In cases involving wearables, obtain the raw sensor data if possible, rather than just the "interpreted" output from the user app. The raw data may be more reliable or may reveal the limitations of the device's inferences.
• Legislative Intent as a Shield: Use the 2012 Parliamentary debates (Minister K Shanmugam's speech) to argue against expansive interpretations of the Evidence Act 1893 that go beyond the "working order" principle.

Subsequent Treatment

As this is a 2025 decision of the Court of Appeal, it currently stands as the definitive authority on the interpretation of s 116A(1) of the Evidence Act 1893. It effectively harmonizes the approaches taken in Telemedia and Super Group, confirming that the "working order" presumption does not extend to substantive veracity. It is expected to be cited in any future case where the reliability of IoT or wearable device data is at issue.

Legislation Referenced

• Criminal Procedure Code 2010 (2020 Rev Ed), s 397(1), s 397(6)(a)
• Evidence Act 1893 (2020 Rev Ed), s 116A, s 116A(1)
• Penal Code (Cap 224, 2008 Rev Ed), s 354(2)
• Australian Commonwealth Evidence Act 1995, Sections 146 and 147

Cases Cited

• Considered: Super Group Ltd v Mysore Nagaraja Kartik [2019] 4 SLR 692
• Considered: Telemedia Pacific Group Ltd v Credit Agricole (Suisse) SA (Yeh Mao-Yuan, third party) [2015] 1 SLR 338
• Referred to: GIL v Public Prosecutor [2024] SGHC 287

Source Documents

• Original judgment PDF: Download (PDF, hosted on Legal Wires CDN)
• Official eLitigation record: View on elitigation.sg