Taiwan

Information uploaded as at September 2026

AT A GLANCE

In Taiwan, AI is used across law enforcement, prosecution, and the judiciary to improve efficiency and support decision-making. Police in several cities have deployed AI-powered patrol systems for license-plate recognition, real-time video, route optimisation, and hotspot monitoring. The National Highway Police use AI to detect seatbelt violations, while facial-recognition tools assist in identifying suspects. For prosecutors, the Ministry of Justice has piloted speech-recognition transcription and an AI Assistance System that can draft documents in drink-driving and fraud cases, schedule fund transfers, flag repeat offenders, and interface with electronic interrogation records. In the courts, the Judicial Yuan has introduced speech-to-text transcription, offline voice-input drafting, intelligent dossier analysis, and a pilot AI judgment-drafting application for certain criminal cases, alongside a publicly accessible AI Sentencing Information System. Private AI legal-research platforms are available to defence lawyers, while the National Police Agency’s 165 Anti-Fraud Dashboard and AI chatbot provide real-time scam consultation for victims. Training for judges and court staff focuses on secure use, data classification, human oversight, and the limits of AI in adjudication.

Despite this scope of use, as at September 2026, Taiwan has no single statute comprehensively regulating AI in criminal proceedings. The Artificial Intelligence Basic Act, passed in 2025, establishes general principles for AI development—including human autonomy, privacy, transparency, fairness, and accountability—and requires government risk assessments, but does not specifically address judicial applications. The Executive Yuan’s 2023 Guidelines on the Use of Generative AI are legally binding internal rules that prohibit use for classified documents, bar decisions based solely on AI output, require security checks before sharing personal information, and mandate disclosure where AI is used. Other existing laws impose indirect constraints through rules on evidence, data collection, cybersecurity, and equality. The Criminal Code criminalises sexually explicit deepfake images, and election law penalises deepfake content relating to candidates, but there is no specific regime governing deepfakes in court proceedings. Taiwanese courts have begun to address AI-related issues in case law, restricting AI-drafted apologies, using generative AI to estimate expenses or market trends, questioning the reliability of facial-recognition evidence without corroboration, and limiting access to courtroom recordings over deepfake concerns, signalling a cautious, human-centric approach to AI integration in the justice system.

USE

In Taiwan, AI tools are used by law enforcement and the courts throughout the criminal process.

Law enforcement

Operational support

Several city police departments have reportedly deployed AI systems to improve operations while on patrol:

In May 2024, the Taichung City Government announced the implementation of ChatGPT to generate press releases. The city stated that the platform can ‘generate press releases of up to 400 to 800 words in less than one minute after inputting simple case conditions’. Taichung Police Chief Li Wenzhang emphasised that the AI system gave rise to an estimated increase in press release writing efficiency by 60% and reduced the rate of human errors.

A new Smart Cloud Identification System was also announced by Taichung City in January 2026. This system integrates AI image analysis, vehicle-mounted video, and street view map data. It provides ‘search by text’ and ‘search by image’ functions, enabling rapid comparison and location identification through street views, signage text, or building exteriors.

A 3D Security Patrol and Early Warning System has been implemented into various areas of Kaohsiung. The system uses an AI algorithm to monitor various hotspots in the city and ‘identify potential security incidents’. The system activates an alarm which instantly notifies police of a potential issue.

Data review and analysis

In October 2024, the Taiwan National Highway Police Department announced the launch of AI violation detection systems in two locations on National Highway 1. The system, called ‘鷹眼辨識’ (‘AI-powered eagle-eye detection technology’), photographs passing vehicles and uses AI software to identify individuals not wearing seatbelts. Following the AI screening, violations are manually reviewed. This AI system has triggered over 5,000 cases in 10 months since it was first implemented.

Taichung City has implemented an AI facial recognition tool to catch individuals illegally posting notices on utility poles. The tool, developed in partnership with Chung Hsing University, uses AI to identify faces and licence plates of offenders and streamline the process to issue fines.

The National Police Agency of Taiwan has used an ‘M-Police Operation System’ to identify individuals using facial recognition technology. The system, procured from a Japanese IT company called NEC, accesses citizen ID photos from a household registration database. It is unclear the extent to which this technology deploys real-time facial recognition technologies. The heightened use of facial recognition technology was evident in the identification of individuals under quarantine during the Covid-19 pandemic to catch violators in populated areas before issuing fines.

The Taiwan Railways Administration has also installed cameras with facial recognition technologies in public transport systems. Such technology is to enhance and expedite response capabilities and provide real-time threat detection.

Prosecutors

Case management

Taiwan’s Ministry of Justice has developed its own speech-recognition platform, piloted in the Taipei District Procuratorate. The process includes the transcription of interrogations, identification of speakers, creating text-searchable transcripts, reducing administrative workloads, and improving efficiency.

Legal research, analysis and drafting support

An AI-powered Assistance System for Prosecutorial Agencies is used by the Ministry of Justice. In May 2024, the Taiwan Judiciary and Legislative Affairs Committee reported that the system has the capability to generate draft documents for drunk driving cases, create schedules of funds transferred by victims in fraud cases, identify repeat offenders, generate recommendations for case referral or dismissal, tag and identify digital case files, and interface with electronic police interrogation records.

Prosecutors also have access to the information available on the Factual Sentencing Information System (see ‘Courts’ section below).

Courts

Case management

In a 2024 survey by the International Association of Judges, Taiwan Judge Jen Sheng Kuo disclosed several AI tools used by Taiwan’s judicial system:

1. Speech Recognition: Taiwan courts use an AI system that relies on Deep Neural Networks to transcribe spoken Mandarin to text in real time. The system was trained on a specialised legal corpus containing publicly available judgements and transcripts. The system is used to transcribe for all participants in a court proceeding or in judicial conferences.
2. Offline Speech Input Software: This AI system uses speech recognition and enables judges and their clerks to use speech to draft judgments or other documents without requiring internet connection. This system also includes a separate extension that allows judges to use voice commands to search for judgments and statutes, or to view electronic case files.
3. Intelligent Analysis System: The Taiwan Judicial Yuan (Taiwan’s judicial authority) uses this system to create electronic case dossiers. The system can analyse case dossiers, create PDF bookmarks for case content, and automatically search and mark various personal data within the dossier for redaction. The system is trained on various litigation documents and reduces the time needed for assistants to organise the materials.

In 2020, the Judicial Yuan conducted a pilot programme that uses AI voice recognition technology to transcribe courtroom proceedings. The programme accommodated Taiwan’s linguistic diversity, transcribing in Mandarin, Taiwanese, and Hakka. The technology was tested first in Taiwan’s High Court as well as district courts in Changhua County, Pingtung County and New Taipei City.

Legal research, analysis and drafting support

In August 2023, Taiwan’s Judicial Yuan announced a pilot programme for an AI application to help judges draft judgments in certain criminal cases. While judges retain full responsibility for fact-finding, legal reasoning, and sentencing, the AI application supports them by generating draft text for the written judgment. The application was developed in collaboration with Chunghwa Telecom and primarily uses the indictment as input to generate a draft judgment. It is built on an MT5 large language model, and trained on indictment and judgment data from 1996 to 2021. Any draft produced must refer to the Executive Yuan’s Guidelines on the Use of Generative AI by the Executive Yuan and its Subordinate Agencies (discussed below) which require human oversight, prohibit use for classified documents, and bar reliance on unverified AI output as the sole basis for official decisions.

The Judicial Yuan has stressed that the system has limited relevance to generative AI in its core function. When extracting facts from the indictment and inserting them into the draft, it uses natural language processing rather than generative AI. Once the judge determines the sentence, the system provides rule-based sample language for the legal application section. Where generative AI components are used, rule-based verification is applied to improve accuracy and reduce errors.

The programme was designed to evolve toward an ‘auto-generate’ option, allowing judges to adjust key metrics and produce a draft decision containing a factual summary, reasoning, applicable legislation, and the verdict and sentence. Judges would select options such as the applicable legal provisions, guilty/not guilty, and confession/denial. Judges would then have to choose whether to adopt the draft judgment and edit accordingly.

The 2023 pilot attracted public and legislative pushback. Critics worried the system could become a ‘guilty judgment generator’ and that training on past judgments might reproduce errors. The original test covered two offence types—dangerous driving (including driving under the influence) and aiding/abetting fraud—chosen because they are relatively simple, single-charge cases. Driving under the influence cases are among the most common criminal matters in Taiwan, accounting for roughly 17.9%–21.4% of all criminal cases in 2020–2021.

In response to concerns, the Judicial Yuan postponed the 2023 pilot and narrowed the scope to driving under the influence cases only, initially limiting testing to the Shilin District Court and Tainan District Court. The pilot was subsequently suspended in October 2023 for reassessment, with further validation planned before any wider rollout.

Decision-making support

The Judicial Yuan has developed an AI Sentencing Information System in order to create more predictability in sentencing decisions. The model was developed in response to the Citizen Judges Act, which requires the involvement of civilian ‘judges’ in the sentencing decisions for certain criminal cases and is designed to guide those civilians in forming their sentencing decision. The system contains two different models, a fact-based model and an evaluation model:

1. 1. The fact-based model uses the facts of the underlying case from the judgment to select sentencing factors that are relevant to the case.
   2. The evaluation model weighs whether the sentencing factors are favourable, unfavourable, or neutral to the defendant.

Defence

As at September 2026, Taiwan has not developed any specific AI tools to assist defence attorneys. Taiwanese companies, however, have developed a number of AI research tools that are widely available for lawyers in the country. This includes, but is not limited to, the following:

• • Lawbot AI is a Taiwanese legal AI platform used to provide legal research, document drafting, contract review and case analysis for legal professionals.
  • Lawsnote AI is a Taiwanese legal AI platform used to provide legal research, AI-powered Q&A with citations, automated contract review, risk identification, and regulatory compliance monitoring for legal professionals.
  • Faleduo is a Taiwan-based generative AI legal assistant to provide legal Q&A, contract review, drafting, and lifecycle management services.
  • TaiLexi AI is a Taiwan-based legal AI tool specialising in legal document analysis and case law research.

Victims

Under Article 319 of the Taiwan Code of Criminal Procedure, victims of crime have direct legal standing and may file a private prosecution against the perpetrator of the alleged crime.

Crime prevention

In August 2024, Taiwan’s National Police Agency released an AI tool called the 165 Anti-Fraud Dashboard. In February 2026, the Agency integrated an AI Chatbot into this dashboard for customer services, providing real-time scam consultation, risk identification, response guidance, and suspicious website verification.

TRAINING

In recent years, Taiwan’s judiciary has taken gradual but deliberate steps to equip judges, court staff, and auxiliary judicial personnel with the knowledge required to understand and responsibly use AI tools. Rather than pursuing full automation or delegating decision-making to AI systems, the Judicial Yuan (司法院) has framed AI primarily as an assistive technology—one that may enhance efficiency, information retrieval, and consistency, while preserving human judgment as the cornerstone of adjudication. As a result, AI-related training within the court system has emphasised technical literacy, information security, risk awareness, and human oversight, rather than advanced programming or machine-learning development.

Training for judges and court staff

AI training in Taiwan’s courts is not governed by a single nationwide mandatory curriculum. Instead, it currently takes the form of institution-led workshops, judge continuing-education lectures, and court-specific pilot programs, often developed in cooperation with academic experts or technology professionals. These initiatives reflect a ‘cautious but proactive’ approach: courts seek to familiarise legal professionals with AI’s capabilities and limitations before expanding its practical deployment.

In April 2026, the Taichung District Court organised a structured workshop titled ‘Information Security Risk Governance Principles and Practical Examples for Court Use of AI Tools’. The training was attended by judges, clerks, and administrative personnel, and was hosted directly by the court president, highlighting its institutional importance. The curriculum focused on secure use of generative AI tools (including enterprise versions of AI systems); prevention of confidential data leakage; and establishment of a ‘three-layer defense model’ for judicial AI governance, consisting of: (i) data classification and access controls; (ii) human decision-making supremacy (AI-outputs as reference only); and (iii) retention of prompt and usage records for accountability. The workshop further introduced practical guidance on using AI for non-substantive tasks, such as document language refinement and issue summarisation, with an explicit recommendation that AI outputs be reviewed by court assistants and confirmed by judges before use.

Beyond one-off workshops, AI education has also been incorporated into judicial continuing education programmes, particularly those aimed at judges. The New Taipei District Court, for example, organised a ‘Judge Mentorship Lecture Series’ in June 2024 dedicated to exploring how AI technologies may be applied in judicial contexts. The lecture was held under the theme ‘When AI Technology is Applied to Judicial Adjudication - Taking Sentencing Prediction, Civil and Criminal Insight Databases, and Mediation of Family and Parental Rights Cases as Examples’ and was personally chaired by the President of the Court, Judge Hsu Shih-Feng. The event brought together approximately 50 judges and judicial assistants from the New Taipei District Court and other courts, reflecting a targeted professional-training format aimed at judicial officers rather than general administrative staff. One such lecture, delivered by an academic expert specialising in AI and social applications, covered: basic concepts of AI and machine learning; practical judicial applications, including AI-assisted sentencing information systems; risks of algorithmic bias, opacity, and over-reliance; and the ethical necessity of retaining human discretion and interpretive authority. This training was explicitly framed as capacity-building rather than endorsement, encouraging judges to understand AI outputs critically rather than treating them as neutral or authoritative.

A central impetus for AI training in Taiwan’s courts has been the introduction of the AI-assisted Sentencing Information System (量刑資訊系統), developed and maintained under the supervision of the Judicial Yuan. Although the system itself relies primarily on natural-language processing to organise and retrieve sentencing patterns from past judgments, its introduction has necessitated targeted training for judges. For example, the Judges Academy’s official curricula for successive pre-service training programs for selected judges include a three-hour course on ‘Sentencing Practice, Including Practical Operation of the Sentencing Information System’. A training session offered by the New Taipei District Court related to this system covered the following topics:

• How the system uses AI to identify and organise statutory aggravating and mitigating circumstances and other sentencing factors drawn from prior judgments;
• How the system structures past sentencing information to help legal professionals examine relevant factors and comparable judgments. The system generates sentencing predictions and estimates the relative weight of sentencing factors; and
• The controversies and challenges involved in the use of AI in sentencing.

As Taiwan has integrated AI systems into court systems across the country, judges have cautioned that AI is an ‘auxiliary tool’ that does not replace a judge’s ‘professional judgment’ but that it may reduce repetitive tasks and allow judges to devote additional time and effort to their cases.

Training for lawyers

As at September 2026, there are no public records available for training initiatives for lawyers on the responsible uses of AI in Taiwan.

Training for law enforcement

According to public reporting, the Taipei City Police Department held a ‘dedicated course’ on ‘AI image analysis, source tracing, misinformation verification tools, and digital footprint analysis’. The course, held in June 2026, was intended to improve the police’s capacity to verify deepfake videos and other manipulated media.

REGULATION

As at September 2026, Taiwan has introduced one dedicated AI law and has issued legally binding rules dictating the use of generative AI by the Executive Yuan and subordinate agencies. These regulations are the documents most applicable to regulating the use of AI in Taiwan’s criminal justice system. Beyond those two documents, Taiwan’s other existing statutes and regulations including the Attorney Regulation Act, the Criminal Procedure Code, and the Personal Data Protection Act, among others, may be construed to regulate or constrain AI use by participants in criminal proceedings.

AI Regulations

AI Basic Act (2026)

In January 2026, Taiwan issued the AI Basic Act, which sets out Taiwan’s national framework governing the development of AI. The Act was initially drafted by the National Science and Technology Council (‘NSTC’), was approved by the Executive Yuan, and in December 2025 was passed by the Legislative Yuan. The Act applies broadly to the administrative authorities, private enterprises, and research institutions engaging in AI research, development, and application in Taiwan. Oversight is decentralised, with each ministry developing sector-specific regulations or guidelines relevant to their domain.

The Act does not make explicit reference to the recognition of human rights. Located in Article 4, the governing principles include:

1. Sustainability and well-being;
2. Human autonomy;
3. Privacy and data governance;
4. Cybersecurity and safety;
5. Transparency and explainability;
6. Fairness and non-discrimination; and
7. Accountability.

In particular, the Act notes that the government should ‘avoid the application of AI that infringes on people’s lives, bodies, [or] freedoms’. The Act also emphasises that when the government uses AI, it ‘should conduct risk assessments and plan risk response measures’.

The Act places the NSTC as the competent authority overseeing AI development. It also established the National AI Strategy Special Committee to coordinate national AI affairs. The Act does not address the use of AI within the judiciary.

Guidelines on the Use of Generative AI by the Executive Yuan and Its Subordinate Agencies (2023)

The Guidelines on the Use of Generative AI by the Executive Yuan and Its Subordinate Agencies were formally issued by the Executive Yuan in October 2023. The Guidelines are legally binding internal rules for the Executive Yuan and its subordinate administrative authorities for facilitating and regulating the use of AI. The Guidelines contain several notable restrictions:

• Article 3: Confidential documents must be drafted personally by the responsible officer, and the use of generative AI is prohibited for these purposes.
• Article 4: Officers cannot provide AI with confidential and personal information without permission and are prohibited from consulting generative AI with potentially personal information. The deployment of generative AI is allowed after confirming the security of the AI system and the classification of confidential documents.
• Article 5: Prohibits individuals from basing decisions solely on materials generated by AI.
• Article 6: Appropriate disclosure is required for departments using AI to assist in executing tasks or providing services.

Guidelines for practitioners

As at September 2026, there are no AI-specific guidelines for practitioners in Taiwan. Taiwan has not formally adopted the UNESCO Guidelines for the Use of AI Systems in Courts and Tribunals (2025). However, broader professional guidelines continue to apply.

Attorney Regulation Act (1925)

Taiwan’s Attorney Regulation Act imposes general professional obligations on lawyers, including duties of competence, diligence, integrity, and honesty toward the court. While these rules do not specifically mention AI, they could be construed to require lawyers to verify AI-assisted research for accuracy and to avoid misleading the court with AI-generated hallucinations or fabrications and be responsible for overseeing the AI in generation of materials.

Judges Regulation Act (1923)

Taiwan’s judges are governed by the Judges Act and related ethical codes, which require independence, impartiality, and professional competence. The regulation does not specifically address the usage of AI, and we have not identified Taiwan authority specifically construing the Judges Act or judicial ethics rules to regulate judicial use of AI. However, these existing duties could provide a basis for limiting judicial reliance on AI tools. First, the duty of independent adjudication may require that judges retain ultimate responsibility for legal reasoning and outcomes rather than deferring to AI-generated recommendations; second, the duty of impartiality may require judges to evaluate whether AI outputs introduce bias or unequal treatment; and third, the duties of competence and confidentiality may require judges to understand the limitations of any AI tool used and avoid entering non-public case information into systems that are not approved for such use.

Criminal procedure rules

Criminal Code (as amended in 2025)

In 2023, the Taiwan legislature enacted legislation criminalising the creation or distribution of sexually explicit AI deepfake images and videos:

• Article 319-4 of the Taiwan Criminal Code, imposes a prison sentence of up to five years on an individual who ‘intends to distribute, broadcast, deliver, display publicly, or use other means to allow others to view the victims’ synthetic sexual images generated using computer synthesis or other technological methods[.]’
• Article 165 of the Criminal Code also states that a ‘person who forges, alters, destroys, or conceals evidence in the criminal case of another or makes use of such forged or altered evidence shall be sentenced to imprisonment for not more than two years[.]’ This particular crime may apply to the use of AI technology to create or alter evidence through deepfakes.

Criminal Procedure Code (as amended in 2020)

The Taiwan Code of Criminal Procedure (‘TCCP’) does not specifically address the use of AI at any stage of criminal procedure. However, the following provisions may be relevant to the use of AI in criminal proceedings in Taiwan:

• Article 165-1 of the TCCP permits the use of ‘electronic record[s] or other similar evidential items’ as evidence and requires the party presenting the evidence to ‘identify’ or ‘explain[]’ the essential points of the record which may be applicable where AI is used to acquire evidence in the proceeding. Decisions by the courts are also subject to specific requirements.
• Article 310 requires that a guilty ruling contain facts relevant for the verdict, the evidence favourable to the defendant, the reasons for the sentence, and the applicable law. Judgements drafted by AI would be required to abide by the same parameters.

Data protection legislation

Personal Data Protection Act (as amended in 2025)

The Personal Data Protection Act (‘PDPA’) is Taiwan’s principal law related to the protection of personal data. The PDPA regulates the ‘collection, processing and use of personal data’ whether by government or non-government agencies. Article 5 of the PDPA generally establishes that the ‘collection, processing and use’ of the data must be carried out in a ‘good-faith manner’, not ‘exceed the necessary scope’, and have a ‘legitimate’ purpose for collection. In particular, the PDPA bars the collection of data related to criminal records unless the purpose falls under one of the exceptions, such as for the government agency to perform ‘its statutory duties’, the data is public, or the data is necessary for statistics gathering or academic research.

The PDPA differentiates between requirements for government agencies and non-government agencies. Generally, government agencies are permitted to collect or process personal data where the data is ‘within the necessary scope’ of its statutory duties, ‘consent has been given’, or the ‘rights and interests of the data subject will not be infringed’. The PDPA also requires government agencies to inform the data subject when collecting, processing or using their personal data. Notably, the Act does not expressly contemplate the collection or use of data in training AI systems; however, depending on the data used to train the AI systems deployed in Taiwan’s criminal justice system, the PDPA may apply.

Cybersecurity laws

Cyber Security Management Act (2021)

The Cyber Security Management Act (‘CSMA’) establishes a comprehensive cyber security regime for public and certain private entities. The CSMA establishes varying levels of cyber security responsibility depending on the level of risk associated with the data held by the entity. For example, an entity may be considered Level A (the highest risk level) if the entity’s ‘business involves the maintenance operation of information and communication systems commonly used for nationwide people services’. An entity may be Level B if the entity has the possession of the archives of personal information of regional or local people.

Under the CSMA, the Judicial Yuan is responsible for approving the cyber security responsibility levels of its subordinate agencies. Government agencies are required to develop and implement a ‘cyber security maintenance plan’ annually that complies with the baseline requirements for the responsibility level that has been assigned to the agency.

Presidential and Vice Presidential Election and Recall Act (as amended in 2026)

In order to counter the rise of deepfake videos relating to political candidates, in 2026 Taiwan amended the Presidential and Vice Presidential Election and Recall Act by raising the term of imprisonment for ‘disseminating, broadcasting or distributing . . . deep-fake voice, image, or electronic recordings of candidates’ to no more than seven years.

Human rights

Constitution of the Republic of China (Taiwan) (1947)

The Constitution of the Republic of China is the supreme law of Taiwan. Several provisions of the constitution may be relevant to the development and use of AI tools in the Taiwanese criminal justice system.

1. Article 7: Article 7 states that ‘[a]ll citizens of the Republic of China, irrespective of sex, religion, race, class, or party affiliation, shall be equal before the law’. This provision may inform Taiwan’s regulation of AI systems, particularly where they are used by governmental actors or are subject to implementing legislation or sector-specific rules addressing discrimination or biased outcomes.
2. Article 8: Article 8 provides a broad guarantee of freedom. It states that ‘[n]o person shall be tried or punished otherwise than by a law court in accordance with the procedure prescribed by law’. This provision may relate to any use of AI throughout the criminal justice process, including use by the courts. Article 8 similarly prohibits an arrest that is outside the ‘procedure prescribed by law’ which may implicate any AI tools used to facilitate the arrest of individuals.
3. Article 12: Article 12 provides that people ‘shall have freedom of privacy of correspondence’. While this is not a broad guarantee of privacy, it may inform Taiwan’s broader legal and policy approach to privacy protection, including in the context of large-scale data collection and use for training AI models.

Outlook

Taiwan has generally indicated a strong desire to integrate AI into its government and society, albeit while maintaining a human-focused view on development and innovation. For example, in 2024, the Executive Yuan announced the creation of a Taiwan-specific Large Language Model, called TAIDE, in order to create a generative AI chatbot with ‘Taiwanese features and traditional Chinese character support’. The press release stated that the government was focused on remaining ‘competitive within the field of generative AI’ to help ‘turn Taiwan into an “AI island”’.

Taiwan will likely continue to integrate AI into its criminal justice system. In May 2025, the Ministry of the Interior announced the establishment of the ‘Research Center for Intelligent Technology and Law Enforcement’. According to the Ministry, the Center will ‘integrate emerging technologies with law enforcement applications, focusing on areas such as big data analytics, AI-driven tools, virtual currency flow analysis, and drone deployment’. This focus on AI tools has already been seen through the rollout of numerous AI tools for policing over the last several years.

CASES

As at September 2026, there are no reported judicial decisions directly addressing the use or misuse of AI by legal professionals. There have, however, been cases concerning deepfakes and procedural issues related to AI, and facial recognition technology.

Deepfakes

In 113 Tai-Shang-Zi-1728 (Supreme Court of Taiwan, May 2024), the Taiwan Supreme Court upheld the Taiwan High Court’s five-year sentence of a YouTuber named Chu Yu Chen for his creation of pornographic deepfake videos. The Court affirmed that Chu violated the Personal Data Protection Act by editing the likenesses of 199 individuals into pornographic videos. The case sparked the amendment to the criminal code prohibiting the creation or transmission of similar deepfake content.

In 111 Sheng 39 (Taiwan High Court, July 2022), though predating the AI Basic Act, the Court restricted party access to copies of courtroom audio recordings, reasoning that once such recordings leave court control, there is no realistic way to stop bad actors from using deepfake technology to fabricate false audio or video using someone’s voice or likeness. The court noted that existing monetary penalties for misuse are only after-the-fact remedies and cannot prevent the recordings from being copied or leaked in the first place.

Procedural rulings

In 114 Shen-Jian-2294 (Taoyuan District Court, April 2026), a judge handling a domestic violence matter ordered the defendant to compose a written apology of at least 2,000 Chinese characters. The court specifically required that the statement be handwritten by the defendant and expressly prohibited any use of AI drafting tools, signalling judicial concern that machine-generated apologies would undercut the sincerity the court was trying to compel.

In 113 Jian-Shang 65 (New Taipei District Court, October 2025), consisting of a personal injury claim against a bus operator, the plaintiff sought reimbursement for taxi costs incurred while traveling to medical treatment. Rather than relying solely on receipts or testimony, the court independently used ChatGPT to estimate what a reasonable taxi fare would be for that route and used the AI-generated figure to support its conclusion that the claimed expenses were legitimate.

The case 112 Shang-Geng(1) 27 (Taiwan High Court Kaohsiung Branch Court, June 2024) centered on a failed land sale intended for pig farming. To assess whether the buyer had suffered genuine financial harm from losing access to the property, the court turned to ChatGPT to research historical trends in pig feed pricing, using that information as one data point in evaluating the damages claim.

In 113 Yi 383 (Nantou District Court, September 2024) a defendant, who was accused of posting content that damaged the complainant’s sexual reputation online, claimed the posts were meaningless text randomly generated by ChatGPT after he entered innocuous prompts like ‘love’ or ‘soap opera’, and that he had published them only to test engagement. The prosecutor countered by showing that ChatGPT actually refuses to produce sexually explicit material matching the post’s content, and submitted the platform’s content policies as proof. The court accepted this rebuttal as evidence that the defendant’s explanation was fabricated.

In 114 Ching-Su 182 (Taoyuan District Court, May 2026), a junior-high student asked ChatGPT whether certain behavior from his living-technology teacher seemed normal, effectively documenting his confusion or distress about the conduct in real time. His mother later found and screenshotted the exchange before the original chat log was deleted during a school-level investigation. The court ruled the screenshots admissible, characterising them as objectively generated digital records with a direct connection to the underlying facts, and finding no evidence they had been tampered or obtained improperly.

Facial recognition technology

In 114 Tai-Shang 5166 (Supreme Court, March 2026), the Supreme Court addressed whether facial-recognition analysis of ATM surveillance footage could establish that a fraud defendant was the person captured on camera. The Court raised serious reliability concerns: facial-recognition results can be skewed by lighting, camera angle, and obstructions, are vulnerable to deepfake manipulation, and suffer from an inherent ‘black box’ problem, since the reasoning behind an AI’s match often cannot be inspected or explained. The Court held that facial-recognition output alone is insufficient to identify a defendant and must be corroborated by other reliable evidence.