February 2026 Newsbyte: Artificial Intelligence – Restrictions and Use

For more information, contact:

Brett Cowell
[email protected]

The growing use of generative artificial intelligence (“Gen AI”) in legal practice has prompted courts around Australia to consider how these tools should be used responsibly. Gen AI is becoming more integrated into legal work.  This provides opportunities to improve access to justice and with responsible use, to increase lawyers’ efficiency and work quality.

Recently, the South Australian Courts published Guidelines for the use of Gen AI, offering helpful insights into what responsible use looks like during litigation. These Guidelines provide practical direction for practitioners already using Gen AI across a variety of tasks, including:

• drafting submissions and pleadings
• summarising large volumes of material
• reviewing documentary evidence
• conducting research.

Used appropriately, Gen AI can save time, reduce costs and streamline processes. However, these benefits must be balanced against a lawyer’s professional and ethical obligations. The Guidelines accept that use by lawyers of Gen AI is inevitable and may be beneficial.  They explore in some detail the tension between benefits of use and a lawyer’s obligations.  The Guidelines appear to us to steer a course that acknowledges the appropriate use by lawyers of Gen AI and the professional obligations they have.

Examples of appropriate use of Gen AI

Refinement of Submissions

A lawyer drafts interlocutory submissions containing no confidential material or suppressed content, employs a Gen AI tool solely to refine wording, then carefully confirms the revisions do not alter the evidence or legal principles. Satisfied the changes are purely stylistic and improve clarity, the lawyer files the document under their own name.

Preparation of Affidavit Evidence

A lawyer preparing an affidavit for a key employee uploads the parties’ discovery material into a confidential Gen AI tool to generate a chronology of relevant communications, using it only to guide their interview preparation. The lawyer then conducts a recorded interview, with consent, using open questions and has Gen AI produce a draft narrative statement based solely on the answers. Carefully reviewing the Gen AI‑generated draft against the recording, the lawyer edits it to ensure it accurately reflects the evidence without embellishment or alteration. The revised document is then marked as a draft and provided to the employee for their own thorough review to confirm it truthfully and accurately represents their evidence.

Review of Transcript

A lawyer uses a Gen AI tool to review the trial transcript and extract the evidence given by each witness on specified topics. They then carefully check each Gen AI‑generated summary to ensure it accurately reflects the witness’s actual testimony, making any necessary corrections. The verified summaries are incorporated into a schedule or table, complete with transcript page references, for use in preparing closing submissions.

Examples of inappropriate use of Gen AI

Elaboration of Submissions

A lawyer drafts an outline of submissions addressing the relevant facts, legal issues, legislation, and authorities, but then uses a Gen AI tool to obtain a generic summary of the legal principles and inserts it directly into the document. The added material is not critically reviewed and includes numerous references of uncertain relevance, transforming the outline into an unnecessarily long submission that imposes additional work on both the other party and the Court.

Preparation of Affidavit Evidence

A lawyer preparing an affidavit for an employee in a workplace dispute improperly relies on Gen AI by uploading either the entire discovery set or the CEO’s draft statement and asking the Gen AI to produce a draft statement on the employee’s behalf. The lawyer then sends this Gen AI‑generated statement to the employee to review. This approach is inappropriate because it creates a witness statement derived from documents or another person’s account rather than from the witness’s own recollection, risking inaccuracy, contamination or fabrication of evidence and undermining the integrity of the witness’s testimony.

Alteration of Evidence

A lawyer has been provided with a blurry CCTV image it wishes to adduce as evidence. The lawyer uploads the image into a multimodal Gen AI system with the prompt to ‘enhance and clarify the image to make the registration plate more readable’. The Gen AI fills in the details which were not discernible in the original image and the Gen AI image is lodged as evidence with no disclosure of Gen AI use. Because Gen AI fills in missing details through probabilistic inference, the resulting image is at least partly fabricated. Presenting such material to the Court without disclosure is a serious violation of the duty of honesty and may amount to an offence.

General Summary

As the application of Gen AI in legal practice continues to expand, a consistent theme emerges from the Guidelines, Gen AI is not a substitute for legal skill or professional judgment. Gen AI is a tool which is effective when used well to support efficiency and improve legal processes. However, lawyers must remain diligent, ensuring Gen AI enhances their work without compromising ethical and professional standards, including their duties to the courts.

For more information, contact:

Brett Cowell
[email protected]

The growing use of generative artificial intelligence (“Gen AI”) in legal practice has prompted courts around Australia to consider how these tools should be used responsibly. Gen AI is becoming more integrated into legal work.  This provides opportunities to improve access to justice and with responsible use, to increase lawyers’ efficiency and work quality.

Recently, the South Australian Courts published Guidelines for the use of Gen AI, offering helpful insights into what responsible use looks like during litigation. These Guidelines provide practical direction for practitioners already using Gen AI across a variety of tasks, including:

• drafting submissions and pleadings
• summarising large volumes of material
• reviewing documentary evidence
• conducting research.

Used appropriately, Gen AI can save time, reduce costs and streamline processes. However, these benefits must be balanced against a lawyer’s professional and ethical obligations. The Guidelines accept that use by lawyers of Gen AI is inevitable and may be beneficial.  They explore in some detail the tension between benefits of use and a lawyer’s obligations.  The Guidelines appear to us to steer a course that acknowledges the appropriate use by lawyers of Gen AI and the professional obligations they have.

Examples of appropriate use of Gen AI

Refinement of Submissions

A lawyer drafts interlocutory submissions containing no confidential material or suppressed content, employs a Gen AI tool solely to refine wording, then carefully confirms the revisions do not alter the evidence or legal principles. Satisfied the changes are purely stylistic and improve clarity, the lawyer files the document under their own name.

Preparation of Affidavit Evidence

A lawyer preparing an affidavit for a key employee uploads the parties’ discovery material into a confidential Gen AI tool to generate a chronology of relevant communications, using it only to guide their interview preparation. The lawyer then conducts a recorded interview, with consent, using open questions and has Gen AI produce a draft narrative statement based solely on the answers. Carefully reviewing the Gen AI‑generated draft against the recording, the lawyer edits it to ensure it accurately reflects the evidence without embellishment or alteration. The revised document is then marked as a draft and provided to the employee for their own thorough review to confirm it truthfully and accurately represents their evidence.

Review of Transcript

A lawyer uses a Gen AI tool to review the trial transcript and extract the evidence given by each witness on specified topics. They then carefully check each Gen AI‑generated summary to ensure it accurately reflects the witness’s actual testimony, making any necessary corrections. The verified summaries are incorporated into a schedule or table, complete with transcript page references, for use in preparing closing submissions.

Examples of inappropriate use of Gen AI

Elaboration of Submissions

A lawyer drafts an outline of submissions addressing the relevant facts, legal issues, legislation, and authorities, but then uses a Gen AI tool to obtain a generic summary of the legal principles and inserts it directly into the document. The added material is not critically reviewed and includes numerous references of uncertain relevance, transforming the outline into an unnecessarily long submission that imposes additional work on both the other party and the Court.

Preparation of Affidavit Evidence

A lawyer preparing an affidavit for an employee in a workplace dispute improperly relies on Gen AI by uploading either the entire discovery set or the CEO’s draft statement and asking the Gen AI to produce a draft statement on the employee’s behalf. The lawyer then sends this Gen AI‑generated statement to the employee to review. This approach is inappropriate because it creates a witness statement derived from documents or another person’s account rather than from the witness’s own recollection, risking inaccuracy, contamination or fabrication of evidence and undermining the integrity of the witness’s testimony.

Alteration of Evidence

A lawyer has been provided with a blurry CCTV image it wishes to adduce as evidence. The lawyer uploads the image into a multimodal Gen AI system with the prompt to ‘enhance and clarify the image to make the registration plate more readable’. The Gen AI fills in the details which were not discernible in the original image and the Gen AI image is lodged as evidence with no disclosure of Gen AI use. Because Gen AI fills in missing details through probabilistic inference, the resulting image is at least partly fabricated. Presenting such material to the Court without disclosure is a serious violation of the duty of honesty and may amount to an offence.

General Summary

As the application of Gen AI in legal practice continues to expand, a consistent theme emerges from the Guidelines, Gen AI is not a substitute for legal skill or professional judgment. Gen AI is a tool which is effective when used well to support efficiency and improve legal processes. However, lawyers must remain diligent, ensuring Gen AI enhances their work without compromising ethical and professional standards, including their duties to the courts.

For more information, contact:

Almaida Askandar
[email protected]

Currently, Indonesia has no law of a statute, which specifically titled with the term of Artificial Intelligence (“AI”). The existence of provisions related to AI are scattered across numerous sectoral regulations in Indonesia. Noticing the growing needs to provide general regulatory provisions in relation to AI, this article will briefly analyze the regulations that specify definition of AI, utilization of AI, and the relevant limitations or restrictions.

I.  General Definition of AI

Based on Chapter I (paragraph 1.3-point c) under the Attachment of Presidential Regulation Number 95 of 2018 on the Electronic-Based Government System, AI means artificial intelligence technology in machine that has cognitive functions to learn and solve problems the same way as performed by human. Furthermore, in accordance with Article 5 (a) of Circular Letter of Minister of Communications and Digital Affairs No. 9 of 2023 on Artificial Intelligence Ethics, AI means a form of programming on a computer device to conduct the processing and/or organizing data in an accurate manner.

II.  Utilization of AI

• Standard Business Classification Code or Klasifikasi Baku Lapangan Usaha Indonesia (“KBLI”)

Based on KBLI Year 2020 which is regulated under Regulation of Head of Statistic Central Bureau No. 2 of 2020 on Standard Business Classification Code and the official website of Online Single Submission of Ministry of Investment and Downstream Activity, the programming activity of the basis of AI is classified under KBLI Code 62015. The explanation of such classification covers the activities of consultation that is followed by the analysis and the programming which utilize the technology of AI including a subset of AI such as machine learning, natural language processing, expert system, and another subset of AI.

In line with the above, Minister of Communications and Digital Affairs Regulation No. 15 of 2025 provides that programming activities forming the basis of AI fall under KBLI Code 62015, which carries the same scope of classification activities.

• Risk-Based Business Licensing

The programming activity on the basis of AI falls into the classification of business licensing for the operational sector of electronic information and transactions (“EIT”) according to Article 186 (1) (d), Paragraph 21 on the operational sector of EIT under Government Regulation Number 28 of 2025 on Risk-Based Business Licensing (“GR 28/2025”). Based on Article 186 (2) of GR 28/2025, Business Licensing to Support Business Activities or Perizinan Berusaha untuk Menunjang Kegiatan Usaha (also known as PB-UMKU) in the operational sector of EIT covers the supporting operations and/or business activity commercial which consist of:

• the classification of game product, and
• the registration of electronic system operation of the private sector, for the entire business sector.

• Mandatory Provisions

Minister of Communications and Digital Affairs Regulation No. 15 of 2025 stipulates the mandatory provisions for KBLI 62015 as follows:

• Creating and implementing the internal company policies on the internal data and ethics of AI.
• Complying with the self-declare for expert worker in accordance with the Standard of Indonesia National Work Competency at least for the Category of Information and Communication of Programming Activity Main Group, Computer Consultation and Other Related Activities in the field of AI Expertise, the sub-field of Data Science.
• Publicizing the technological innovation and development to public through event, demonstration, and other methods which can be accessed by public while having the attention to the aspects of privacy and information legality.
• Reporting the activities of data security and governance activities to the MCDA annually.

• The Intersection of AI and the Processing of Personal Data

In condition where the utilization of AI is correlated with the personal data, then Law No. 27 of 2022 on Personal Data Protection (“PDP Law“) is applicable. Article 1 (1) of PDP Law defines Personal Data as data regarding an identified or identifiable individual, either individually or in combination with other information, either directly or indirectly, through electronic or non-electronic systems. Within this framework, AI is not treated as a legal subject under Indonesian law and, as such, cannot assume legal responsibilities under the PDP Law. Nevertheless, any personal data processing carried out through AI is legally ascribed to the party that defines the purpose of the processing and maintains control over it.

Consequently, the party that operates or deploys the AI system is regarded as the Personal Data Controller and bears full legal accountability for compliance with the PDP Law. In this capacity, the Personal Data Controller must safeguard and ensure the security of the Personal Data being processed by: (a) establishing and implementing operational and technical measures to prevent disruptions in processing that violate applicable laws and regulations; and (b) determining an appropriate level of Personal Data security by considering the characteristics and risks inherent in the Personal Data involved in the processing activities.

• AI Ethics and Responsibility

Referring to Circular Letter of the Minister of Communications and Digital Affairs No. 9 of 2023 on Artificial Intelligence Ethics, several ethical and responsibility guidelines are established in relation to the use of AI. Under the section on Implementation and Responsibility, the implementation of Artificial Intelligence must be carried out on the basis of applicable ethics and codes of conduct governing Business Actors and Electronic System Providers (Penyelenggara Sistem Elektronik/PSE). Furthermore, the operation of AI technology must observe AI ethical values, which include inclusivity, humanity, security, accessibility, transparency, credibility and accountability, personal data protection, sustainable development and environmental considerations, as well as intellectual property.

• Indonesia’s National Strategy on AI 2020 – 2045 (“National Strategy”)

Indonesia has developed a National Artificial Intelligence Strategy, prepared by a Working Group formed by the Agency for the Assessment and Application of Technology (Badan Pengkajian dan Penerapan Teknologi/BPPT). Although the National Strategy does not constitute a binding regulatory instrument, it outlines priority sectors and key programs for AI implementation, including ethics and policy, talent development, infrastructure and data readiness, research, and industry innovation, as well as sectors such as healthcare, bureaucratic reform, education, food security, and smart cities. One of its key initiatives is the issuance of a Presidential Regulation on AI, intended to create legally binding obligations for stakeholders to achieve the Strategy’s targets. As of October 2025, this regulation has been drafted and completed and is currently being finalized, with one notable provision requiring AI products to bear an AI label.

III.  Restriction in Relation to the Utilization of AI

• • AI Must Not Infringe Third-Party Copyright

Under Law No. 28 of 2014 on Copyright (“Copyright Law”), copyright grants exclusive economic rights to the Author or Copyright Holder. Pursuant to Article 9 paragraph (1) of the Copyright Law, such economic rights include, publication of the works; reproduction of the works in all forms; translation of the works; adaptation, arrangement, or transformation of the works; distribution of the works or their copies; performance of the works; communication of the works; and rental of the works. Further, Article 9 paragraph (2) of the Copyright Law stipulates that any person exercising such economic rights must obtain prior authorization from the Author or Copyright Holder. Article 9 paragraph (3) of the Copyright Law expressly prohibits any person, without such authorization, from reproducing and/or commercially using a copyrighted work.

In this context, the utilization of AI systems does not diminish or override the exclusive economic rights granted under Article 9 of the Copyright Law. Where AI systems reproduce, adapt, transform, or otherwise process copyrighted materials, whether as training data or in generating outputs, such acts may fall within the scope of economic rights protected under the Copyright Law. The automated nature of AI processing does not eliminate the legal qualification of such acts as reproduction or transformation of a work. Therefore, the use of AI systems involving copyrighted materials requires prior authorization from the Author or Copyright Holder.

• The Prohibition in Connection with the Utilization of Personal Data

In condition where the utilization of AI is correlated with Personal Data (as mentioned in Section II.D above), Article 65 of PDP Law states the prohibition provisions in Personal Data utilization as follows:

• Every person is prohibited from unlawfully obtaining or collecting Personal Data that does not belong to him/her with the intention of benefiting himself/herself or other, which may cause the loss to the subject of Personal Data.
• Every person is prohibited from unlawfully disclosing the Personal Data that does not belong to him/her.
• Every person is prohibited from unlawfully using the Personal Data that does not belong to him/her.

For more information, contact:

Liew Sue Yin
[email protected]

First reported use in court proceedings

In Malaysia, use of artificial intelligence was introduced in the courts in the states of Sabah and Sarawak (situated in East Malaysia) in January 2020 with the launch of the Artificial Intelligence in Court Sentencing system (AiCOS). AiCOS analyses data and patterns for sentencing purposes. The intended purpose in its deployment is to assist the judiciary in maintaining consistency in sentencing decisions.

It was used in February 2020, in criminal proceedings conducted in the magistrates court for offences under the Dangerous Drugs Act 1952, to recommend appropriate sentencing for the offender, for consideration by the magistrate. 

In Peninsular Malaysia (or West Malaysia), it was reported in 2025 that the courts are also exploring the use of AI in sentencing guidelines.

AI as a tool for efficiency in the legal and judicial landscape

In Malaysia, AI tools are used by the courts for case management, legal research and administrative efficiency.

The current judicial philosophy in Malaysia is that the use of AI is to enhance judicial efficiency but not as a replacement for human judgment or at the expense of justice.

Limitations of use in the legal and judicial landscape

One of the recognized limitations of AI as a tool is that it is not able assess human factors such as the credibility of a witness in a trial or take into consideration intangible elements relevant for purposes of context. These limitations spotlight the unique nature of the human factor notwithstanding advances in technology.

From the perspective of professional practice –  Using it judiciously and responsibly

The Malaysian Bar Council has from time to time, issued circulars addressing the use of AI as a tool with a key emphasis on responsible use of AI as a tool and its impact on existing obligations of practitioners with respect to ethical standards, client confidentiality and duties to clients, courts and other stakeholders. 

The Malaysian Bar Council does not endorse or regulate the use of any form of generative AI (including Chat GPT).

Members of the legal profession have been advised of the importance of verifying the output of AI generated work product and that AI should not be the sole source of the legal preparation process of any work product. Members of the legal profession have also been cautioned to ensure that they continue to adhere to ethical standards, client confidentiality and their professional duty to clients, the courts and other stakeholders at all times.

The national policy perspective

From a national policy perspective, there is currently an initiative to drive the adaptation of AI technology into the country’s legal and judicial infrastructure. This initiative involves the participation of the Legal Affairs Division of the Prime Minister’s Department, the Office of the Chief Registrar of the Federal Court of Malaysia, the Legal Aid Department, the Malaysia Department of Insolvency, the National Law Academy, the Human Rights Commission of Malaysia, and the Asian International Arbitration Centre as well as collaboration with the Ministry of Digital, the Ministry of Communications, the Attorney General’s Chambers, the Malaysian Bar Council and state legal bodies. While this initiative is specific to the legal and judicial landscape, there are also other AI initiatives taking place at the national level.

At the broader national level, it was reported in November 2025 that the Ministry of Digital aims to position Malaysia as a leading AI nation where Malaysia is expected to emerge as an inclusive, ethical, safe and competitive AI nation, where AI is a key driver of economic growth and transformation of citizen’s lives by the year 2030.

Malaysia’s National AI Office (“NAIO”) is tasked with amongst others, developing policies and regulatory frameworks, fostering innovation, accelerating AI adoption across all sectors, building talent development and ensuring that AI developments are ethical and inclusive. The NAIO is an agency under the Malaysia Ministry of Digital.

The Malaysia Digital Economy Corporation (MDEC), also government agency under the purview of the Ministry of Digital, is charged with leading Malaysia’s digital economy. It has launched initiatives focused accelerating Malaysia’s digital economy including supporting the development and commercialization of AI solutions by Malaysia Digital companies.

For more information, contact:

Dan Hughes
[email protected]

Artificial intelligence in New Zealand – restrictions and uses 

• The New Zealand Government endorses an overarching “light-touch, proportionate and risked-based” approach to AI use and regulation. This is intended to be a balanced approach, avoiding the risks of perceived over-regulation, such as in the case of the European Union’s AI Act; or the risks of a largely free-market, laissez-faire approach, as seen in the United States. 
• The Minister of Science, Innovation and Technology, Rt Hon Judith Collins KC, has set out for New Zealand to be a “fast follower” of AI development, whereby future AI regulation will be informed by closely observing global trends, alongside learning from other countries’ regulatory successes and failures.

Use of AI in New Zealand

• As part of its “light-touch” approach, the Government has established several guidance frameworks for the use and uptake of AI in the private and public sectors.
• In July 2025, the New Zealand government released its Strategy for Artificial Intelligence (AI Strategy). Broadly, the AI Strategy aims to encourage AI adoption in the private sector and reinforce businesses’ obligations under existing legal frameworks.  One component of the AI Strategy is the Ministry of Business, Innovation and Employment’s ‘Responsible AI Guidance for Businesses’ framework which gives practical tips to assist organisations with safely and efficiently implementing AI in their everyday work. 
• Alongside the AI Strategy, the Government has also issued its ‘Responsible AI Guidance for the Public Service’ which seeks to enable responsible AI adoption and use within public sector agencies to drive productivity and economic growth in New Zealand.
• Viewed together, these frameworks exemplify the Government’s preference to provide soft guidance on AI use in the private and public sectors, as opposed to implementing AI-focused legislation.

Restrictions on the use of AI in New Zealand

• Currently, there are no laws per se in New Zealand restricting the use of AI. Rather, the Government has amended existing legislation to provide what it considers to be necessary restrictions on AI use. 
• An example of such an amendment is the passing of the Biometric Processing Privacy Code. The Code, which amends the Privacy Act 2020, governs organisations’ collection and use of biometric information, including by AI, which is processed through biometric technologies. 
• The Privacy Act 2020 (Act) is the most significant legislative instrument governing restrictions on AI use. The Act applies to any agency (including overseas entities carrying out business in New Zealand) that collects, holds or processes personal information.  Personal information also encompasses inaccurate or false information about an individual, such as deepfake images or fake social media profiles.   The Act is technologically neutral, meaning that it applies to AI systems that process data about identifiable individuals and can be easily amended without substantive redrafting to address AI-related issued.
  • The Act is comprised of 13 information privacy principles (IPPs). The IPPs apply to the use of AI tools in collecting, using and sharing personal information:
    • IPP1: Consider whether inputting personal information into the AI tool is directly connected to the initial purpose for which the information was collected.
    • IPP2: Agencies must generally obtain personal information directly from the person concerned.
    • IPP3: Agencies must be transparent about the collection of information, such as how and why it will be utilised.
    • IPP4: Data must be collected through a lawful and fair process.
    • IPP5: Agencies must proactively assess potential security risks and prevent unauthorised access to personal information.
    • IPP6-7: Organisations must establish guidelines on how it will process requests from individuals to access and edit their own personal information.
    • IPP8: Steps must be taken to ensure the accuracy of outputs, particularly in light of the limitations and gaps within AI’s capabilities.
    • IPP10-11: Identify the purpose for the collection of information and restrict its use and disclosure to this purpose and any such directly related aims.

Potential future developments on the use and regulation of AI in New Zealand

• However, despite the Government’s stated intentions, the possibility of tailor-made AI legislation is not entirely off the table. One such example is the Deepfake Digital Harm and Exploitation Bill 2025, introduced to Parliament in October 2025.  This bill seeks to amend the Crimes Act 1961 and the Harmful Digital Communications Act 2015 by expanding the definition of “intimate visual recording” to explicitly include images created, synthesised, or altered to show a person’s likeness produced without consent.  These proposed changes primarily focus on combatting the rise in production and dissemination of deepfake pornography; as well as providing further legal protection for victims.  Given the recent controversy surrounding X’s (fka Twitter) Grok AI image tool, which has been used to produce and spread deepfake pornography, this bill has become increasingly prescient in highlighting the dangers of inadequate, or even non-existent, regulation of AI tools. 
• In these early days, the Government’s “light-touch” and “fast-follower” approach seems to be a sensible and pragmatic policy initiative to incentivise and balance the risks of AI use in New Zealand. However, given the exponential growth and advancement of AI technologies, it remains to be seen how the Government’s approach will address the issues posed by rapid and unprecedented technological development in the long-term. 

For more information, contact:

Sham Chee Keat
[email protected]

Appreciating the super-accelerated evolution of AI, the Singapore government intentionally refrains from pre-maturely formulating a single comprehensive set of AI law, which may be subjected to overly frequent changes, and which may stifle the growth of AI.

Instead, the Singapore government takes on a pragmatic and proactive approach. It collaborates with AI industry leaders and chooses to formulate soft AI governance frameworks and guidelines to promote the adoption and use of AI, adjusting the same as and when it deems necessary, and at the same time constantly reviewing and amending various areas of laws that are closely related to AI or the use thereof.  In other words, instead of playing catch-up reactively, the Singapore government prefers to lead the way together with the major industry players in shaping the research and deployment of AI.

The following AI initiatives (non-exhaustive) have been put forward by the Singapore governments in recent years:

• National AI Strategy (2019 version);
• Singapore launched the world’s first Model AI Governance Framework in 2019 (First Edition, January 2019);
• Model AI Governance Framework (Second Edition, January 2020);
• AI Verify, an AI governance issued in 2022, a testing framework and software toolkit being made open source for developers in June 2023;
• National AI Strategy 2.0 (2023 version), which sets out the ideals towards AI to transform cognitive and physical tasks, to steer AI for the public good, to unlock the next frontier of economic growth and to address risks from the potential abuse and mismanagement of AI;
• Model AI Governance Framework for Generative AI (MGF for GenAI) (January 2024); and
• Model AI Governance Framework (MGF) for Agentic AI (January 2026), the first of its kind in the world. It sets out the risks, alerting AI users to the human accountability in AI use, risk assessment and control.

The existence of the government’s frameworks and guidelines has wide-ranging legal implications. One’s  failure to meet the standards set by the frameworks or guidelines may eventually be seen as one’s breach of a duty of care with legal consequences in say the performance of contracts, in tort or in compliance-related issues.

Many harms potentially associated with AI are already addressed by existing laws. At the same time, Singapore has updated existing laws to address gaps created by using AI.  Below are some examples.

In the Singapore courts, the Supreme Court Registrar’s Circular No. 1 of 2024 was issued on “Guide on the use of generative AI tools by court users” which sets out general principles to regulate the use of AI by court users (both lawyers and litigants).  The court does not prohibit the use of Generative AI tools to prepare court documents, provided that its guide is complied with, without changing a court user’s duty to comply with the relevant legislation, rules, codes of conduct and practice directions. Whether the court user is a lawyer or a litigant in person, he or she has to ensure that materials put before the courts are independently verified, accurate, true, and appropriate.

The Personal Data Protection Commission has published advisory guidelines on the use of personal data in AI recommendation and decision systems, providing clarity on how personal data can be used to deploy or develop machine learning AI models.

The Protection from Online Falsehoods and Manipulation Act 2019 empowers the government to issue correction directions to contents creators and providers for fake news or statements circulated online that are against the public interest, regardless of whether they are generated by AI.  

The guidelines for fair employment practices impose on employers the responsibility to recruit and select employees based on merit, and to guard against bias on the grounds of age, race, gender, religion, marital status and family responsibilities, or disability, even if any bias was generated by the use of AI.

In the area of copyright, Copyright Act 2021, repealing the old Act, was re-enacted to introduce, amongst others, a new defence to copyright infringement for machine learning to promote research and innovation.

In October 2024, the Intellectual Property Office of Singapore released the Singapore Guidance for the Examination of AI-related Patent Applications, which sets out the legal criteria for patent eligibility in Singapore in AI-related patent claims.

The Penal Code 1871, which governs the criminal law, was amended in 2019 to include the offence of “sextortion”, where one threatens or distributes an intimate image of a victim, regardless of whether AI has been used.  The existing Prevention  Protection from Harassment Act 2014 protects a victim from being victimised by AI generated images of him or her that may cause him or her harassment, alarm or distress.

“This is a living document. We have worked with government agencies and leading companies to collate current best practices, but this is a fast-developing space, and best practices will evolve. This framework will need to be continuously updated to keep pace with new developments.”  Quoting from the MGF for Agentic AI 2026, this is where Singapore is, constantly promoting AI in safe, living and responsible frameworks while addressing the risks of any abuse or mismanagement of AI.

For more information, contact:

Weita Liao
[email protected]

Taiwan’s Artificial Fundamental Act (the “AI Act”) was passed and formally promulgated and be put into force in January 2026. As Taiwan’s first foundational statute with legal authority specifically addressing artificial intelligence (the “AI”), the AI Act adopts a human-centered approach as its core principle. The AI Act is intended to promote the application of AI technologies and the development of related industries, safeguard the fundamental rights guaranteed under the Constitution, and advance sustainable social development, while ensuring digital inclusion and equity.

Legal Characterization and Fundamental Principles of the AI Act

For purposes of the AI Act, “artificial intelligence” refers to systems capable of autonomous operation that, through data input or sensing and by means of machine learning, algorithms, or related technologies, generate outputs—such as predictions, content, recommendations, or decisions, aimed at achieving explicit or implicit objectives and capable of affecting physical or virtual environments.

As a principles-based statute, the AI Act does not replace or supersede existing Taiwanese laws governing personal data protection, information security, healthcare, financial services, or other regulated industries. Rather than serving as a sector-specific or technology-specific regulatory regime, the AI Act is positioned as a declaratory and overarching framework statute. Its purpose is to establish the foundational structure for the development, use, and governance of AI in Taiwan, clarify the allocation of governmental responsibilities, and provide a legal basis for the formulation and implementation of subsequent subsidiary regulations and policy measures.

Additionally, the AI Act expressly articulates seven fundamental principles governing the research, development, and deployment of AI, which serve as reference benchmarks for government authorities when formulating policies, issuing guidelines, or adopting related regulatory or administrative measures:

1. Sustainability and Social Welfare: Appropriate education and training should be provided to mitigate potential digital divides.
2. Human Autonomy: AI systems should respect fundamental human rights, including personality rights and cultural values, and be designed and deployed in a manner that ensures meaningful human oversight and intervention where appropriate.
3. Privacy Protection and Data Governance: Personal data privacy and trade secrets should be protected, data minimization principles should be adopted, and the openness and reuse of non-sensitive data should be promoted.
4. Cybersecurity and Safety: Adequate cybersecurity safeguards should be established to prevent security threats and attacks.
5. Transparency and Explainability: AI-generated outputs should be subject to appropriate disclosure or labeling.
6. Fairness and Non-Discrimination: Risks of algorithmic bias and discriminatory outcomes should be minimized to the extent practicable.
7. Accountability: Responsibility attribution and liability conditions should be clearly defined, and appropriate remedial mechanisms should be established.

Accordingly, the AI Act primarily operates through the foregoing high-level principles to facilitate cross-sectoral coordination in the development and application of AI technologies.

Regulatory Scope and the Role of Government Authorities of AI Act

The regulatory focus of the AI Act is primarily placed on the Taiwanese government and the competent authorities of various regulated sectors. The AI Act expressly assigns to such authorities the responsibility to promote the development of AI and to establish corresponding governance measures. These responsibilities include, among others, fostering international cooperation; developing AI risk classification frameworks aligned with international standards for the purpose of implementing risk-based regulatory approaches; and, with respect to the deployment of high-risk AI systems, defining the allocation of responsibilities and applicable regulatory conditions, as well as establishing mechanisms for remedies, compensation, or insurance. In addition, the AI Act requires each competent authority to review the laws, regulations, and administrative measures within its respective remit and, where any such provisions are inconsistent with the objectives or principles of the AI Act, to complete the necessary review and, as appropriate, undertake amendments, repeals, or improvements within two years following the effective date of the AI Act.

As the AI Act itself does not establish any penalty provisions, it does not, at this stage, impose direct and specific legal obligations on private entities or enterprises. Nevertheless, the AI Act elevates governmental measures relating to subsidies, tax incentives, and investment incentives for AI-related industries to the statutory level and incorporates institutional mechanisms for data sharing and data reuse. These measures are intended to reduce structural barriers to the development of AI technologies by enterprises. When supporting regulations and implementing measures are progressively adopted, such mechanisms are expected to facilitate access to data resources and address practical constraints commonly faced by enterprises, including limitations in data volume for the training of high-quality AI models. Accordingly, the risk management and accountability principles articulated under the AI Act have already had tangible, albeit indirect, impacts on enterprise AI research, development, and use in Taiwan.