RE: PROPOSED GOVERNANCE FRAMEWORK ON THE USE OF HUMAN-CENTERED AUGMENTED INTELLIGENCE IN THE JUDICIARY

Sirs/Mesdames:

Please take notice that the Court En Banc issued a Resolution dated February 18, 2026 which reads as follows:

“A.M. No. 25-11-28-SC

RE: PROPOSED GOVERNANCE FRAMEWORK ON THE USE OF HUMAN-CENTERED AUGMENTED INTELLIGENCE IN THE JUDICIARY

RESOLUTION

WHEREAS, a key technological advancement in the 21^st century is the rapid and unprecedented progress in artificial intelligence (AI) breakthroughs affecting nearly all aspects of life and society;

WHEREAS, the global acceleration of AI research and development with the corresponding growth of associated industries has extended to the proliferation of the use of AI in the justice sector, for legal practice, and in courts;

WHEREAS, when AI is developed, deployed, and used without the appropriate human oversight and control, and inconsistent with human values and human rights, the technology causes harm, such as amplification of preexisting biases, inequalities, discrimination, and widening of the digital divide; propagation of misinformation, disinformation, and malinformation; and the acceleration of the climate emergency;

WHEREAS, notwithstanding emerging and evolving technological advancements, the Supreme Court of the Philippines (Supreme Court) remains committed to upholding justice and the rule of law;

WHEREAS, the Supreme Court recognizes the potential of human-centered augmented intelligence, as a form of AI, to enhance the administration of justice and make court operations more efficient, while acknowledging technology’s inherent limitations and the risks involved in its use;

WHEREAS, the ethical and responsible use of human-centered augmented intelligence tools in the Judiciary should be consistent with the principles of transparency and accountability to reinforce the public’s faith and confidence in the independence and impartiality of the judicial system, and to affirm the duties and responsibilities of members of the Judiciary, court officials and employees, and members of the Bar, respectively, under ethical rules and regulations, including but not limited to the 2025 Code of Judicial Conduct and Accountability (CJCA), the 2025 Code of Conduct and Accountability for Court Officials and Personnel (CCACOP), and the Code of Professional Responsibility and Accountability (CPRA);

WHEREAS, the Supreme Court En Banc, in A.M. No. 22-04-26-SC, crafted the Strategic Plan for Judicial Innovations for 2022–2027 (SPJI 2022–2027), which lays down the reform initiatives necessary toward achieving a technology-driven Philippine Judiciary (Judiciary), including the use of human-centered augmented intelligence to improve processes in the Judiciary, taking into account current laws and regulations and available technology and infrastructure;

WHEREAS, in 2023, the Council of Association of Southeast Asian Nations (ASEAN) Chief Justices Working Group on Facilitating Service of Civil Processes Within ASEAN presented the Governance Framework on the Use of Artificial Intelligence for the ASEAN Judiciaries, and in 2025, the United Nations Educational, Scientific and Cultural Organization (UNESCO) released its Guidelines for the Use of AI Systems in Courts and Tribunals;

WHEREAS, in line with the Innovation Outcome of the SPJI 2022–2027, the Supreme Court has introduced pilot applications for human-centered augmented intelligence in the Judiciary, such as voice-to-text transcription of stenographic notes;

WHEREAS, under Article VIII, Section 5(5) of the Constitution, the Supreme Court is vested with the power to “promulgate rules concerning the protection and enforcement of constitutional rights, pleading, practice, and procedure in all courts, the admission to the practice of law, the Integrated Bar, and legal assistance to the underprivileged”;

WHEREAS, acknowledging the need for guidelines to achieve responsible use of human-centered augmented intelligence in the judicial system based on three core ethical guidelines: (1) fairness; (2) accountability; and (3) transparency, the Supreme Court through its Committee on Computerization and Library in Memorandum Order No. 13-2025 constituted a Working Group for the Creation of a Governance Framework on the Use of Artificial Intelligence in the Judiciary, composed of the following:

Chairperson	Hon. Marvic M.V.F. Leonen Senior Associate Justice, Supreme Court
Vice Chairpersons	Hon. Ramon Paul L. Hernando Associate Justice, Supreme Court Hon. Rodil V. Zalameda Associate Justice, Supreme Court
Members	Hon. Apolinario D. Bruselas, Jr. Associate Justice, Court of Appeals Hon. Marian Ivy F. Reyes-Fajardo Associate Justice, Court of Tax Appeals Hon. Pia Cristina B. Bersamin-Embuscado Regional Trial Court Judge Hon. Niño Delvin E. Embuscado Regional Trial Court Judge Atty. Jed Sherwin G. Uy Chief Technology Officer, Supreme Court Atty. Maria Victoria Gleoresty Sp. Guerra Chief Attorney, Supreme Court Atty. Miguel Martin A. Buenaventura Office of the Chief Justice Atty. Michelle S. Duguil Office of Associate Justice Alfredo Benjamin S. Caguioa
Consultants	Dr. Prospero C. Naval University of the Philippines Diliman, Department of Computer Science Atty. Senando Angelo R. Santiago University of the Philippines College of Law, Information and Communications Technology Office

NOW THEREFORE, the Supreme Court En Banc resolves to ADOPT the following Governance Framework on the Use of Human-Centered Augmented Intelligence in the Judiciary (AI Governance Framework, or this Framework).

February 18, 2026, Manila, Philippines.

A.M. No. 25-11-28-SC

GOVERNANCE FRAMEWORK ON THE USE OF HUMAN-
CENTERED AUGMENTED INTELLIGENCE IN THE PHILIPPINE
JUDICIARY

RESOLUTION

PART I
GENERAL PROVISIONS

A.

PURPOSE

Consistent with the SPJI 2022–2027, the Supreme Court aims to innovate and use modern technologies, such as human-centered augmented intelligence, to enhance operational efficiency and expand access to justice.

The Supreme Court recognizes both the benefits and risks associated with the use of human-centered augmented intelligence. It thus aims to provide a framework that will guide the Judiciary, the Bench, and the Bar in upholding ethical standards, fostering accountability, and safeguarding societal welfare, while promoting technological innovation and creativity.

This Framework shall undergo periodic review to ensure that it is responsive and relevant to the technological advancements and evolving best practices in AI governance for the judicial system and courts.

B.

SCOPE

The AI Governance Framework shall apply to the following in their use of human-centered augmented intelligence tools as defined in Part I.C.6:

1. Members of the Judiciary;
2. court officials and employees;
3. court users; and
4. vendors, third-party contractors, or organizations engaged in the design, development, and utilization of AI tools by or on behalf of the Judiciary.

C.	DEFINITIONS OF TERMS

1. Artificial intelligence (or AI system^[1] refers to any computational system designed for or has the effect of processing data to perform tasks typically associated with human intelligence, including aspects of reasoning, learning, perception, prediction, planning, or control.^[2] AI may be standalone systems or integrated in computer software, applications, or other tools.
2. Human-centered augmented intelligence refers to AI that is developed, deployed, and used consistent with the ethical principles and in accordance with the management governance of this Framework.
3. Human-centered augmented intelligence tool (“AI tool”) refers to any software or hardware, or combination thereof, which wholly or partially uses, integrates, relies on, or operationalizes any AI technology, which is developed, used, or controlled by a human being.
4. Machine learning refers to a subdomain of human-centered augmented intelligence which broadly pertains to the development of computers that learn patterns from data or perform specific tasks with increasing improvement without being explicitly programmed.[3]

5. Large language models, in AI, refers to a type of AI system that processes human language, through training data of large quantities of texts.[4]

6. Use of human-centered augmented intelligence refers to any set of operations, including the creation, design, development, implementation, and deployment of AI systems, or tools, in whole or in part, whether directly or indirectly, as a means of accomplishing a purpose or achieving a result, by:
   1. members of the Judiciary, and court officials and employees, in the performance of their official functions, including but not limited to adjudication, administration, rulemaking, or allied purposes;
   2. court users in all their dealings and submissions with the courts or other offices in the Judiciary, including the preparation and submission of court-bound documents, and presentation of evidence in any matter; and
   3. vendors, third-party contractors, or organizations, acting on behalf of or for the benefit of the Judiciary or any member thereof, for any adjudication, administration, or allied purpose.
7. Members of the Judiciary refers to justices and judges of all court levels.
8. Court officials and employees refer to all court officials and employees, regardless of status of employment with the Judiciary.
9. Court users refer to members of the public, including members of the Philippine Bar and Shari’ah Bar (collectively, members of the Bar), who seek relief from, transact, or interact with the courts or offices of the Judiciary, whether on their own behalf or for another person or entity.
10. Stakeholders refer to the following: 
    1. members of the Judiciary, and court officials and employees;
    2. members of the Bar;
    3. legal educational institutions and legal academe;
    4. vendors, third-party contractors, or organizations engaged in the design, development, and utilization of AI tools by or on behalf of the Judiciary;
    5. court users and the public;
    6. academics, researchers, and scientists in the field of AI and the law; and
    7. the justice sector and government agencies engaged in AI policymaking and regulation.

PART II
ETHICAL GOVERNANCE

1.	HUMAN-CENTERED

The use of AI must be human-centric, maximizing the potential of the technology to improve the efficiency, accuracy, and capabilities of human users, while minimizing the identified and unanticipated risks that such use may create and avoiding the amplification of existing harms to individuals, groups, or society at large.

The use of human-centered augmented intelligence should be centered on human values, such as the promotion of the rule of law and fundamental freedoms, dignity and autonomy, privacy and data protection, fairness, nondiscrimination, and social justice.

Sustainable human development requires individual and communal capacity-building where technology works towards solving human problems and addressing human needs. All persons must be aware of the effects of indiscriminate and unmoderated use of AI and consumption of its outputs on cognitive ability,^[5] critical thinking,^[6] human interpersonal relationships and community-building.^[7] Therefore, human-centered augmented intelligence tools should primarily be utilized as a support or augmentation tool that complements and empowers human cognitive skills, rather than completely supplanting or overriding human judgment.

2.	RESPECT FOR HUMAN RIGHTS AND HARM AVOIDANCE

All use of technology within the Judiciary must be consistent with the Constitution and international human rights instruments to which the Philippines is a party. The use of human-centered augmented intelligence tools must be anchored on the rule of law, as well as the protection and promotion of human rights and fundamental freedoms.

In case the use of an AI tool or its output has the effect of limiting human rights and fundamental freedoms, the use must have a legal basis, and must not go beyond what is reasonable, proportional, and necessary to achieve a legitimate aim. In all cases, the method used must be the least intrusive or restrictive means to achieve a legitimate purpose.

However, regardless of the benefit or potential of any particular AI tool, if there is evidence that such tool presents an equivalent or even greater harm, or probability of such harm, to any entity or stakeholder, in violation of their rights or contrary to the rule of law, then any member of the Judiciary, court official or employee, or court user shall not use such AI tool.

The development and use of an AI tool or any of its components, such as datasets, shall also respect the intellectual property rights and creations of human artists, creators, scientists, and inventors.

3.	TRANSPARENCY

Notwithstanding specific technical complexities, all use of human-centered augmented intelligence should be understandable and intelligible to the public. The capabilities and risks associated with human-centered augmented intelligence must not be obscured by reason of apparent or actual complexity of the technology employed. To this end, enhancing the technological literacy of all members of the Judiciary, court officials and employees, and court users and informing the public about advancements in AI that the Judiciary is using or intends to use must be prioritized.

The use of human-centered augmented intelligence in the Judiciary must be explainable as to both the process and the outcome. The Judiciary must inform the public of useful information about how AI tools are developed, trained, and deployed in the Judiciary, without jeopardizing the confidentiality of court proceedings, security of judicial operations, and standards of legal and judicial ethics.

In case a member of the Judiciary, or court official or employee uses an AI tool or any output thereof in any process in the Judiciary, such use must be clearly disclosed and explained in plain and understandable language to cultivate public trust and confidence. At minimum, such disclosure must clearly indicate the following: (1) the AI tool or tools used including its/their version, and the purpose for using it/them; (2) the extent or degree of use of the AI tool/s and/or their output in a particular work. or task; (3) the extent or degree of human control and oversight in relation to the particular work or task where the AI tool was used; (4) a statement that in case of inquiry or upon request, the user preserved the output of the AI tool/s; (5) compliance with the relevant and applicable principles found in this Framework; and (6) a statement that the user bears ultimate responsibility for the work or output done, in accordance with the principle of accountability. The disclosure shall be in accordance with guidelines to be issued by the Court.

Disclosure shall be mandatory and in clear and plain language accessible to all parties, when members of the Judiciary, and court officials and employees use human-centered augmented intelligence tools in the preparation of court-issued documents in the exercise of their adjudicatory functions, for purposes of, including, but not limited to the following:

1. voice-to-text transcription;
2. translation;
3. automated compilation or generation of structured authorities, citations, or other paratext;
4. legal research;
5. summarization of documents;
6. automated document processing, including optical character recognition;
7. copy-editing or proofreading; and
8. redaction or sanitization of data from documents in compliance with law, rules, or regulations.

The datasets, algorithms, and processes that result in the AI tool’s outputs should be comprehensively documented to facilitate traceability.

Developers of AI tools procured or intended to be procured by the Judiciary shall be required to disclose the tool’s logic, limitations, and safeguards undertaken to ensure transparency and accountability. In the procurement of any AI tool wholly or partially owned, developed, created, or designed by any third party, it must fully and completely disclose to the Judiciary the nature, processes, data models used, and other significant components of the tool.

Any invocation of trade secrets, confidentiality, or proprietary information must be justified if it may prevent or limit disclosure on the nature, processes, data use and processing, or results of the tool.

4.	ACCOUNTABILITY, HUMAN OVERSIGHT, AND CONTINUOUS MONITORING

Human control must be paramount in any use of AI. While human-centered augmented intelligence aims to approximate aspects of human intelligence and cognition, it must always be understood that this should not replace human discernment.

Responsibility ultimately falls on the designer, developer, or user of an AI tool. As such, a user of an AI tool—whether a member of the Judiciary, a court official or employee, or court user—is personally responsible for the output the tool produces and its consequences.

An AI tool shall neither remove nor diminish the control, management, and supervision by human beings. Notwithstanding any automated or semi-automated process of the said tool, all outputs must be reviewed and approved by human beings. As such, the use of human-centered augmented intelligence shall neither excuse the responsibility of the user for the consequences of such use, nor become a ground to mitigate or justify any liability or penalties imposable under law or codes of ethical conduct.

Under no circumstance should AI tools or their output serve as the sole, primary or determinative basis of any adjudicatory outcome. Legal reasoning and final conclusions that determines the rights and duties of parties before courts or any Judiciary office must be independently formed by the human decision-maker.

A member of the Judiciary, court official or employee, or a member of the Bar shall not evade responsibility for any violation of law, regulation, or code of conduct including but not limited to the CJCA, CCACOP, or CPRA, or for any breach of legal or judicial ethics, with a claim that the violation or breach was the fault of or caused by an AI tool.

Human-centered augmented intelligence tools used in the Judiciary must be auditable^[8] and traceable,^[9] and systems must be put in place to assess, manage, and minimize the associated risks. Both the AI tools and the methods and means by which these tools are monitored, assessed, and evaluated must undergo periodic review to determine their effectivity in light of the current state of technology and evolving norms in the use of AI.

The Court shall continuously monitor the use of human-centered augmented intelligence by all courts and offices within the Judiciary to determine their effectivity and efficiency in achieving the goals for which the tool is used, and their alignment with the principles enumerated in this Framework.

Local, regional, and global developments in AI technology for the justice sector shall be monitored to ensure that the Judiciary has access to the best available systems, technology, and tools that align with its fundamental principles and strategic plans.

5.	FAIRNESS AND NON-DISCRIMINATION

Users of AI tools must recognize that these are made and used by humans who are prone to conscious and unconscious bias. Users must likewise be aware of (a) the biases engendered by new forms of technology, especially those that use natural language or mimic human-like conduct; (b) training models of such tools which may have used datasets that are incomplete, which may overvalue or undervalue certain demographics, or which are inappropriate for the unique contexts for which the tools are to be deployed; and (c) the use of AI tools could also exacerbate existing biases against vulnerable, underrepresented, and marginalized populations.

To guard against these possibilities, the development and use of human-centered augmented intelligence in the Judiciary must not only be free from overt discrimination of any form, but must also be protected against perpetuating existing inequalities or creating new forms of discrimination. The use of AI tools must be guided with respect, sensitivity, and inclusivity across diverse identities, roles, needs, and vulnerabilities.

The Judiciary shall promote capability-building programs against algorithmic bias, automation bias, and other cognitive fallacies and negative effects associated with human-computer interactions.

6.	PRIVACY AND DATA PROTECTION

The Judiciary processes data in all aspects of its operations. A significant amount of this data is personal information, sensitive personal information, confidential data, and those kept secret by reasons of public safety, national security and defense, and other lawful grounds. The right to privacy must be protected, and data protection must be promoted from the design of an AI tool until its use. Users of AI tools must be aware of what data is entered into the tool, how the data is processed by the tool, and who has access to the data used by the tool. Any use of an AI tool must be consistent with data privacy regulations and judiciary-issued data governance policies. No AI tool shall be used to process confidential, privileged, sensitive, or secret information without express authority and under the strict supervision of the Court, or any authorized office thereof.

7.	SECURITY, SAFETY, AND ROBUSTNESS

The Judiciary’s enterprise information systems may leverage AI to enhance security of the Judiciary’s information technology assets at all stages, from threat detection and assessment, to secure encryption of communications and records, and to defend against threat actors.

However, human-centered augmented intelligence may also be used to attack and undermine Judiciary assets and personnel in increasingly sophisticated ways. Thus, AI tools must be used under a preventive approach to risks and unacceptable harm.

A comprehensive risk assessment is a necessary prerequisite before the implementation of any AI tool for any reason in any part of the Judiciary. Moreover, the risk assessment must be explained in plain language to the end-users of the AI tool so that they may intelligently and cogently understand the risks associated with their use.

AI tools, like all other information technology used by the Judiciary, should be resilient to attacks like data poisoning, model leakage, or attacks against both software and hardware that may lead to changes in data or system behavior.

Thus, safety measures should be in place to minimize the risks of unintended consequences. AI tools must be continuously classified according to their risks, in accordance with the system risk models of this Framework. In particular, the use of predictive AI, and the outputs thereof must be assessed as a high-risk system, because of the consequences arising from inaccurate or inadequate predictions; thus, it is important to ensure that accuracy pertains to any predictive AI tool’s output based on data or models. When inaccurate predictions are inevitable, the tool must indicate the likelihood of these errors.

8.	SUSTAINABILITY OF RESOURCES AND ENVIRONMENTAL IMPACT

The use of AI tools in the Judiciary should be made sustainable and have minimal or no negative impact on the environment. In balancing innovation and sustainability, preference must be given to the most energy-, water-, and resource-efficient technology in addressing a problem.

The Judiciary shall ensure that energy or water consumption during the use of AI tools stay within a set appropriate range and be reduced when necessary.

Before any AI tool is used, there must be reliable evidence that it will achieve its intended purpose, and safeguards have been built in the tool which would justify its use in relation to possible ecological harm or effect. If such evidence or safeguards are lacking, or when there is uncertainty as to the environmental impact an AI tool may cause, the precautionary principle shall apply. An AI tool shall not be used if it will cause or result in serious, irreversible, or disproportional damage to the environment.

PART III
MANAGEMENT GOVERNANCE

Change requires management attention at the highest level. A new process is more likely to be accepted when all stakeholders are involved. Management governance proposes a process that enlists top management participation and the involvement of operational management. This section outlines seven main actions aimed at operationalizing the principles set out in the preceding section:

1.	ORGANIZATIONAL AND LEADERSHIP STRUCTURE
A.	IN GENERAL

All members of the Judiciary, and all court officials and employees, have a shared responsibility in the use of human-centered augmented intelligence.

The Supreme Court En Banc, in the exercise of its judicial power, rule-making power, and administrative supervision over all courts and members of the Bar, shall oversee and supervise all aspects of AI—including human-centered augmented intelligence—policies and processes in the Judiciary, the Bench, and the Bar. It shall set policies, through rules and guidelines, on the responsible uses of human-centered augmented intelligence and shall determine and approve, on a case-by-case basis, the allowable extent of utilization of AI tools by the persons falling within the scope of this Framework, as enumerated in Part I.B.

The Management Information Systems Office (MISO) shall advise the Supreme Court En Banc on the technical aspects of the use of human-centered augmented intelligence in the Judiciary.

The Supreme Court data protection officer shall ensure that the Court, as both a personal information controller (PIC) and a personal information processor (PIP),^[10] is able to safeguard data that is utilized in connection with the use of human-centered augmented intelligence in the Judiciary.

The Office of the Court Administrator (OCA) shall implement change management directives issued by the Supreme Court En Banc for the first-and second-level courts, including capacity-building and upskilling programs for court officials and employees.

Similarly, the Philippine Judicial Academy (PHILJA) shall provide avenues for members of the Judiciary, and court officials and employees to undertake certificate training, programs, courses, and other learning opportunities within the field of AI and the law.

Likewise, legal education institutions and the legal academe are encouraged to engage in research and development in and offer courses in the field of AI and the law, including ethics, applications, and regulations. They are further encouraged to adopt measures to specifically impart knowledge on “generative AI tools[‘] . . . adequate uses, limitations, and the risks associated . . . with their adoption for the drafting of legal documents and supporting participation in legal proceedings, as well as supporting court management activities.”^[11]

Meanwhile, the Integrated Bar of the Philippines (IBP), in the interest of elevating the standards of the legal profession and of enabling the Bar to discharge its public responsibility more effectively, shall monitor the legal profession’s engagement with and use of AI tools, and regularly report its findings to the Supreme Court. The IBP is also encouraged to utilize the Continuing Legal Education (CLE) program to create opportunities for law practitioners in the private sector to receive training in the field of AI and the law, including the ethical and responsible use of human-centered augmented intelligence.

B.	THE COMMITTEE ON HUMAN-CENTERED AUGMENTED INTELLIGENCE

A permanent committee, to be known as the Committee on Human-Centered Augmented Intelligence (“AI Committee”) shall be established as an advisory body that will provide leadership and guidance on how AI tools are designed, developed, used, and deployed within the Judiciary. The AI Committee shall also provide direction on how AI tools may be ethically and responsibly used in the practice of the legal profession.

Duties and functions

The duties and functions of the AI Committee shall include the following:

1. recommend to the Supreme Court En Banc policy guidelines on the responsible use of AI in the Judiciary, as defined in Part I.C.6., and by the legal profession, including rules, regulations, or guidance addressing specific types of AI, e.g., generative AI, predictive AI, their implementation for particular functions or processes, or the intersection of AI with legal disciplines or domains;
2. oversee the development and deployment of AI tools within the Judiciary by making recommendations and advisories to the Supreme Court En Banc regarding (a) the setting of the Judiciary’s risk appetite with respect to specific AI tools, on a case-to-case basis, (b) the selection of AI tools to be procured for specific court operations and management activities, and (c) the setting of the extent of participation of third-party developers in the deployment of AI tools procured by the Supreme Court;
3. supervise the procurement of AI tools from third-party providers, in accordance with the ethical principles and procurement policies of this Framework, including formulating procurement assessment guidelines, for the Supreme Court En Banc‘s approval;
4. evaluate emerging AI tools for ethical issues or AI-related risks prior to their deployment within the Judiciary, including the conduct of consultations on developing, updating, and recommending to the Supreme Court En Banc the processes for (a) incident reporting, (b) vulnerability reporting, (c) risk assessment, (d) continuous risk management measures, and (e) impact assessment;
5. promote and regularly conduct capacity-building and management training for Members of the Judiciary, court officials and employees;
6. promote and initiate research and development, including interdisciplinary efforts to foster innovation in applications of AI within the Judiciary, legal practice, and the justice sector;
7. receive and evaluate complaints or incident reports related to the use of AI tools by court officials and court employees, and third-party contractors, or organizations engaged in the use, especially design and development, of AI tools by or on behalf of the Judiciary, and provide recommendations to the Supreme Court En Banc;
8. after evaluation of complaints or incident reports on the use of AI tools in the Judiciary, recommend the temporary or permanent suspension of the use of an AI tool if the continued use presents harm to any entity or stakeholder, in violation of their rights or contrary to the rule of law;
9. provide guidance to the Supreme Court En Banc relative to the disposition of AI-related cases;
10. study the feasibility of developing AI literacy and competency standards for (a) Members of the Judiciary; (b) court officials and employees; (c) lawyers and Shari’ah counselors; and (d) law students and legal educators;
11. coordinate with and issue guidance for other stakeholders on emerging issues in AI and their use, including the promotion and enforcement of the principles of this Framework to the public;
12. create technical working groups for the conduct of risk and impact assessments prior to the deployment of an AI tool and for other purposes; and
13. conduct a periodic review of this Framework, to ensure that it remains responsive and relevant to the current state of technology and best practices in AI governance.

Composition

It is recommended that the Committee be composed, at minimum, of the following:

1. a member of the Supreme Court who shall sit as the Committee Chairperson;
2. a justice from an appellate court;
3. a judge from the first- or second-level courts;
4. a representative from the Supreme Court OCA;
5. a representative from the Supreme Court MISO;
6. a representative from the Supreme Court Data Privacy Office;
7. a representative from the PHILJA;
8. a representative from the IBP;
9. at least two subject matter experts on AI, one of whom must have experience in the legal sector or judicial technology; and
10. a representative from the academe, with expertise in the ethical aspects of technology.

The AI Committee shall have the power to appoint or select consultants or resource persons, who shall be legal, technical, or subject matter experts or those with experience in the industry.

2.	DETERMINATION OF THE LEVEL OF HUMAN INVOLVEMENT IN AI-AUGMENTED DECISION-MAKING

In line with the principles of transparency and human oversight, the Court may apply the following human-control approaches to the use of human-centered augmented intelligence tools depending on the level of autonomy that may be granted to the subject AI tool:

The extent of oversight mechanisms required depends on the AI tool’s application area and potential risk.

3.	THE AI ADOPTION PROCESS

No AI tool shall be used unless authorized by the Supreme Court En Banc.

Any sector or office of the Judiciary, upon approval of the Court En Banc, may use AI tools which (1) may be procured from external providers; (2) may contain or utilize components procured from external providers; (3) may be internally designed or developed, with or without the services of an external provider; or (4) any combination thereof.

These are identified roles in relation to their activity in the stages of any AI tool’s development until end-use:

A.	Procurement of AI tools

Any AI tool, which will be procured without the Court having been involved in its design and development, must be compliant or compatible with the ethical governance principles of this AI Governance Framework.

The Court may adopt a comprehensive ethical impact assessment tool or procedure to screen human-centered augmented intelligence tools for potential procurement.

Moreover, the Court may adopt more stringent pre-procurement testing and validation requirements for human-centered augmented intelligence tools designed and developed without the Court’s input. Such testing may entail adversarial methods, where internal red teams attempt to find and expose system vulnerabilities, and where external actors are brought in, with incentives, to similarly discover, document, and report system errors and weaknesses.

In instances where the Court considers procuring an AI tool with continuous learning practices, it shall adopt rigorous monitoring and testing measures for ensuring that the tool, upon being deployed and fed data from a live environment, is consistently kept compliant and compatible with the ethical governance principles in this Framework.

In connection with this, the Court may adopt phased implementation of AI tools, in which case the Court will start with pilot testing, followed by a staggered rollout.^[20] The gradual rollout will allow for finetuning based on early feedback.

B.	Creation of an AI tool

Where Members of the Judiciary, court officials and employees are involved in the design and development of a human-centered augmented intelligence tool, it is recommended that these be done in a phased manner. The following stages can be considered:

Identification of dataset. This involves determining what datasets to use and what data should be included or excluded when examining it for patterns.^[21] Data providers must ensure that the contents of these datasets are compliant with existing regulations on data protection and are not in violation of the right to privacy or data privacy rights of any persons.

Collection, extraction, and analysis of dataset. Large quantities of data alone are not enough. The data quality also affects the validity, accuracy, and usefulness of the outcomes generated by the algorithm. Bias can arise in algorithms in several ways. First, the data collected may have been preferentially sampled, and therefore, the data sample itself is biased. Second, bias can arise because the collected data reflects existing societal bias. There must be awareness and recognition that pre-existing and widely-used datasets are predominantly foreign or international, and are inadequately localized or representative of the local conditions and demographics, or sufficient for local needs. To the extent that society contains inequality, exclusion or other traces of discrimination, so too will the data. The human-centric approach can include different groups of people with different experiences to address issues related to diversity, race, class, sexuality, and religion.

To mitigate the risks involved with the presence of existing bias in the datasets, AI system creators should adopt the use of heterogenous and/or representative datasets, where data is gathered from various reliable sources. The use of synthetic data^[22] must always be justified, clearly disclosed, and closely monitored. The personnel responsible for labelling data should be provided with clear guidelines to establish an objective and repeatable process for each act of individual labelling decisions.^[23]

Localization must be considered, and depending on the use case, be of the highest priority in the collection of data and choice of dataset. Within this context, data localization refers to designing and conducting data collection, especially in terms of population identification and sampling, in a way that captures the context of the locality where the AI tool may be deployed, especially in terms of language and culture.^[24]

Algorithms and model creation. As a scientific discipline, AI includes several approaches and techniques, such as machine learning (of which deep learning and reinforcement learning are specific examples), machine reasoning (which includes planning, scheduling, knowledge representation and reasoning, search, and optimization), and robotics (which includes control, perception, sensors and actuators, as well as the integration of all other techniques into cyber-physical systems). There are many different machine learning algorithms, and each type of algorithm uses a different set of mathematical techniques to try to determine the “best” and “most predictive” factors. Machine learning models depend on data, and existing societal or structural biases may be subtly embedded in datasets. When AI tools use datasets skewed in various ways, these same biases will be reflected in the machine learning models in ways that are extremely hard to detect but may offer disparate treatment to various societal groups based on inappropriate categories. This must be validated and measured to ensure the model works accurately or fairly. Thus, AI system creators and AI system operators should regularly evaluate AI tools for bias and take steps to minimize cases where particular groups are systematically disadvantaged or advantaged by the AI tools’ output and make corrections when such cases are observed.^[25]

Continuous testing, as a risk management measure. Output may also be biased if the training data does not represent the real-world environment the system is supposed to perform. An algorithm may produce unexpected and undesirable results if it encounters a situation markedly different from the data it was trained on. When applying an algorithm and interpreting its outcome, it should always be verified if the algorithm is likely to produce a reliable result for these circumstances, given the environment in which it was tested.

The first relevant consideration in testing the algorithm is to select the type of algorithm. Next, testing the algorithm involves selecting a test dataset and a training dataset. An algorithm will be trained on the training dataset and will be evaluated against the test set. Both datasets need to be representative of the problem that the algorithm is to solve. If the training dataset contains defects, the same defects should be removed from the testing dataset. A thorough testing regime tests the algorithm against realistic scenarios to ensure it is fit for its purpose.^[26]

Choice models.^[27] As algorithms become more sophisticated, their choice-making process may become less tractable. The problems associated with biases in data and insufficient testing may get more pervasive as the choice model gets more complex because these biases may be more difficult to predict or identify.^[28]

Implementation. When AI tools or their output are used in any process, the AI user should be able to explain how the tool arrived at its result and why the AI user has included the tool’s output in their decision or resolution. In principle, the AI user should be able to determine exactly what happened by examining the input information, the AI model itself (what factors it considered and how it weighed them), and how the AI model treated the information to be able to reconstruct the computational process that led to the results. However, even if the details of the AI tools were accessible, in some cases, AI users still may not have complete transparency as to the decision-making process for technical reasons or interpretability problems. These techniques often encode their patterns in extremely complex mathematical models that are readily processable by computers but, unfortunately, are not comprehensible to humans in meaningful terms, even to AI system creators. The usefulness of transparency may depend on the risk associated with the result. The Court may require transparency for outputs of AI tools that have a greater impact on a person. For lower-priority or minimal-risk use-cases, the Court may opt for less transparency in favor of better system performance.^[29]

Monitoring and evaluation; algorithmic auditing. All stakeholders—from policymakers to AI system creators and operators, to AI users—must be involved not only to determine what the AI tool has to do, but also how the AI tool needs to be measured and evaluated, to ensure that the AI tool consistently does what it is intended to do. Monitoring and evaluation mechanisms should be in place to ensure that AI tools are in consonance with human rights and common values. All these phases should include solutions to prevent, detect, and address algorithmic bias (data biases or design biases) that may arise intentionally or unintentionally. Some design biases might be largely unintentional, such as when AI system creators make a choice based upon a personal judgment without realizing that such a choice benefits people like themselves.

Likewise, there must be measures for determining the need for, and for executing, adjustments to the AI tool, coupled with continuous algorithmic auditing.^[30]

4.	STAKEHOLDER CONSULTATION AND CAPACITY-BUILDING

All justices, judges, and court officials and employees, regardless of rank and nature of appointment, in the Judiciary are equally stakeholders in this Framework and the use of human-centered augmented intelligence in the Judiciary.

Other stakeholders include members of the legal profession and all persons within legal educational institutions, as well as vendors, third-party contractors, or organizations engaged in the development and use of human-centered augmented intelligence technologies by or on behalf of the Judiciary. It must also be acknowledged that the public; academics, researchers, and scientists in the law and AI domain; the justice sector; government agencies, especially those in AI policymaking and regulation are also stakeholders in the Judiciary’s development and use of AI tools, and formulation and implementation of AI policies and rules. The Judiciary is committed to timely and clear communication to the public of developments in AI as used, or is planned to be used, by it, without compromising the confidentiality of court processes, security of judicial operations, and requirements of legal and judicial ethics.

Stakeholder consultation is important to ensure transparency and confidence in the entire process. As the buy-in of different stakeholders is vital, inputs should be solicited through consistent interactions with stakeholders to test, evaluate and review the effectiveness of human-centered augmented intelligence to build trust between stakeholders and the Judiciary.

Moreover, in accordance with the ethical and professional requirement of competence in the discharge of duties, there shall be systematic programming to enhance the technological literacy of members of the Judiciary and court officials and employees, especially justices and judges who in the exercise of adjudicatory powers may encounter novel issues on the interpretation of AI-related laws and regulations; use of human-centered augmented intelligence tools or their outputs by litigants in claims, causes of action, or as evidence in court proceedings; and use of appropriate AI tools or output for the speedy disposition of cases or promotion of inclusivity and accessibility in courtrooms.

Furthermore, the Court shall specifically pursue the development of internal (1) tool monitoring and auditing capacity, to avoid overreliance on external parties for risk assessment, impact assessment, and algorithmic auditing,^[31] and (2) cybersecurity capabilities, in view of the fact that any AI tools used by or in the Judiciary will be prime targets for adversarial attacks.^[32]

5.	RISK MANAGEMENT APPROACH AND REGULAR ASSESSMENTS

The Court should apply a systematic risk management approach throughout the AI system life cycle—with assignment of clear roles and responsibilities, the inclusion of regular impact assessment reviews, and the establishment of a mandatory incident reporting protocol—to identify, assess, and mitigate any potential risks, including digital security, safety, and bias, that can affect the system’s behavior and outcome. Further, members of the Judiciary, court officials and employees should have the capacity to review, question, and override system-generated outcomes, where necessary.

The Court may also prepare contingency procedures which shall be made available to members of the Judiciary, and court officials and employees, in the form of scenario-based response plans, in case risk management efforts fail or fall short of their intended targets.

Further, the Court, following the recommendations of the AI Committee, shall adopt a comprehensive risk management framework, which shall then be localized to suit the needs of the Philippine Judiciary. Any such framework must, at the very least, address the following aspects of risk management:

1. Risk measurement. This covers the identification and classification of risks at each stage of the AI lifecycle; the attribution of risks and the mapping of specific risks’ interplay, especially post-deployment of the AI tool; the tracking of emergent risks; the establishment of risk parameters, metrics, and reporting protocols; and the setting and updating of human baselines.
2. Risk tolerance. This covers organizational readiness assessments; harm/cost-benefit tradeoff analyses; and limit determinations and adjustments.
3. Risk prioritization. This covers the resource allocation for risk management measures; and the establishment of protocols for residual risk management.
4. Whole-of-organization measures. This covers the interoperability and integration of the risk management policies and procedures of the separate offices comprising the Judiciary; the alignment of the Judiciary’s risk management policies with those of the departments and agencies (under the other two branches of the government); and the alignment of the Judiciary’s risk management policies with national AI strategies and roadmaps.

The Court’s risk management framework should also be able to address the risks presented by various kinds of AI, and classify tools and process in accordance with the risks they already present, or may present:

1. Prohibited AI systems. These cover AI systems that must necessarily be banned from use by the Judiciary due to their posing of unacceptable risks to fundamental rights and human safety, e.g., AI for cognitive behavioral manipulation, AI for biometrics-based categorization of people, and AI for real-time biometrics-based identification and tracking of people.
2. High-risk AI systems. These cover AI systems that directly affect fundamental rights and human safety, e.g., AI used in law enforcement.
3. Limited-risk AI systems. These cover AI systems that only indirectly or potentially affect fundamental rights and human safety, e.g., customer-service chatbots, transcription.
4. Minimal-risk AI systems. These cover AI systems that pose little threats to fundamental rights and human safety, if any, such that the latter may be secured by requiring disclosure to end-users that they are interacting with an AI system, e.g., spam filters.

6.	DATA GOVERNANCE

Under the Data Governance Framework of the Judiciary ver. 1, dated March 3, 2025 (Data Governance Framework), to ensure the accuracy, security, privacy, and compliance of data used for the Judiciary’s data governance and analytics journey, the following steps are to be undertaken for its proper enforcement:

1. establishment of data governance policies, which include data classification, data retention, data access, and data sharing;
2. data minimization, to prioritize the use of human-centered augmented intelligence tools that “require minimal personal data to function,” and data anonymization, to be done when the said tools “process personal data, especially in creating databases used for legal analytics or precedent study”;^[33]
3. assign data ownership and responsibilities to ensure accountability for data quality and security;
4. develop data quality standards to ensure accuracy, completeness, and consistency of data;
5. ensure data security and privacy so that only authorized persons can access data and ensure compliance with privacy regulations;
6. establish data architecture and standards to ensure consistency in data storage, integration, and sharing;
7. implement data governance controls and process to ensure policies, controls, and processes—including data monitoring, auditing, classification, tagging, incident response and data breach procedures—are established and implemented; and
8. measure and report on data governance effectiveness.

The AI Committee shall continuously coordinate with the Data Protection Officer on data protection and data privacy policies, especially with respect to (1) guidelines and compliances for data sharing, whether by entering into data sharing agreements (under the framework provided by the National Privacy Commission) or by executing memoranda of agreement (pursuant to a Judiciary-crafted administrative circular); (2) the conduct of privacy impact assessments in relation to Judiciary-deployed AI tools; and (3) the resolution of data subjects’ complaints or requests (in exercise of their rights under the Data Privacy Act of 2012) with respect to judiciary-deployed AI tools.

7.	DOMESTIC AND INTERNATIONAL COOPERATION ON BEST PRACTICES FOR TRUSTWORTHY AND RESPONSIBLE AI DEVELOPMENT, DEPLOYMENT, AND USE

The ethical and responsible use of AI tools in the Judiciary calls for cooperation with domestic, regional, or global fora,^[34] to encourage the sharing of AI knowledge and best practices and disseminate and use comparable metrics to assess the performance of AI systems based on empirical evidence.

To this end, the Court may adopt and localize both principles and best practices from the AI governance frameworks of other jurisdictions and organizations, such as, but not limited to the following:

1. the European Union Artificial Intelligence Act;
2. UNESCO Recommendation on the Ethics of Artificial Intelligence, as well as the Guidelines for the Use of AI Systems in Courts and Tribunals;
3. the Global Toolkit on AI and the Rule of Law for the Judiciary;
4. the European Ethical Charter on the Use of Artificial Intelligence in Judicial Systems and their environment;
5. the European Commission High-Level Expert Group on Artificial Intelligence’s Ethics Guidelines for Trustworthy AI;
6. the Association of Southeast Asian Nations (ASEAN) Guide on AI Governance and Ethics;
7. the Expanded ASEAN Guide on AI Governance and Ethics – Generative AI;
8. the Council of ASEAN Chief Justices (CACJ) Governance Framework on the Use of Artificial Intelligence for the ASEAN Judiciaries;
9. the Council of Europe’s Framework Convention on Artificial Intelligence; and
10. the Organisation for Economic Co-operation and Development (OECD) AI Principles.

Domestically, the Court may pursue and foster cooperation with authorities developing national AI policies, such as, but not limited to the following:

1. the Department of Science and Technology, in relation to the National AI Strategy for the Philippines;
2. the Department of Trade and Industry, in relation to the National AI Strategy Roadmap 2.0;
3. the Department of Information and Communications Technology, in relation to its efforts (with the Civil Service Commission) to draft the Principles and Guidelines for an Ethical and Trustworthy Use of AI in the Government;
4. the National Privacy Commission, for the data protection aspect of AI; and
5. other regulatory or research bodies on AI to be created by law or regulation in the future.

With respect to the academe, the Court may explore cooperation with universities and colleges having AI or law and technology programs, as well as with the Department of Education, in relation to the latter’s Education Center for AI Research, and the Commission on Higher Education.

^[2] UNESCO, Recommendation on the Ethics of Artificial Intelligence, available at https://unesdoc.unesco.org/ark:/48223/pf0000381137 (last accessed on February 20, 2026).

^[3] Wannes Meert, Tinne De Laet, and Luc De Raedt, “Artificial Intelligence: A Perspective From the Field“, in THE CAMBRIDGE HANDBOOK OF THE LAW, ETHICS AND POLICY OF ARTIFICIAL INTELLIGENCE, Nathalie A. Smuha (ed.) (2025).

^[4] Idan A. Blank, What are large language models supposed to model?, 27 TRENDS IN COGNITIVE SCIENCE 11, 987-989 (2023); Emily M. Bender, et al., On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?, FACCT ’21: PROCEEDINGS OF THE 2021 ACM CONFERENCE ON FAIRNESS, ACCOUNTABILITY, AND TRANSPARENCY, 610-623 (2021); Rishi Bommasani, et al., ON THE OPPORTUNITIES AND RISKS OF FOUNDATION MODELS (2021).

^[5] Michael Gerlich, AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking, 15 SOCIETIES 9, 252 (2025); Mathura Shanmugasundaram & Arunkumar Tamilarasu, The Impact of Digital Technology, Social Media, and Artificial Intelligence on Cognitive Functions: A Review, 2 FRONTIERS IN COGNITION (2023).

^[6] Kate Goddard, et al., Automation Bias: A Systematic Review of Frequency, Effect Mediators, and Mitigators, 19 J. AM. MED. INFORM. ASSOC. 1, 121-127 (2012).

^[7] Valerio Capraro, et al., The Impact of Generative Artificial Intelligence on Socioeconomic Inequalities and Policy Making, 3 PNAS NEXUS 6, 191 (2024).

^[8] Auditability here refers to an AI system’s readiness to undergo an assessment, either by internal or external parties, of its algorithms, data and data sets, and design processes. (ASEAN Guide on AI Governance and Ethics, p. 35, available at https://asean.org/wp-content/uploads/2024/02/ASEAN-Guide-on-AI-Governance-and-Ethics_beautified_201223_v2.pdf [last accessed on February 20, 2026].)

^[9] Traceability here refers to the practice of using documentation—of (1) the data gathering process, the raw data collected, the data cleaning, the data grouping and labeling, and the setting of algorithms, and (2) the methods used for testing and validating the algorithms, as well as the test results—to facilitate audits. (ASEAN Guide on AI Governance and Ethics, p. 12, available at https://asean.org/wp-content/uploads/2024/02/ASEAN-Guide-on-AI-Governance-and-Ethics_beautified_201223_v2.pdf [last accessed on February 20, 2026]; Ethics Guidelines for Trustworthy AI, pp. 17, 29, available at https://www.europarl.europa.eu/cmsdata/196377/AI%20HLEG_Ethics%20Guidelines%20for%20Trustworthy%20AI.pdf [last accessed on February 20, 2026].)

^[10] Republic Act No. 10173 (2012), secs. 3(h) and 3(i) provide:

SECTION 3. Definition of Terms – 

(h) Personal information controller refers to a person or organization who controls the collection, holding, processing or use of personal information, including a person or organization who instructs another person or organization to collect, hold, process, use, transfer or disclose personal information on his or her behalf. 
. . . .
(i) Personal information processor refers to any natural or juridical person qualified to act as such under this Act to whom a personal information controller may outsource the processing of personal data pertaining to a data subject.

^[11] UNESCO, Guidelines for the Use of AI Systems in Courts and Tribunals, p. 25, available at https://unesdoc.unesco.org/ark:/48223/pf0000396582 (last accessed on February 20, 2026).

^[12] ASEAN Guide on AI Governance and Ethics (2024), pp. 25-26, available at https://asean.org/wp-content/uploads/2024/02/ASEAN-Guide-on-AI-Governance-and-Ethics_beautified_201223_v2.pdf (last accessed on February 20, 2026); UNESCO Global Toolkit on AI and the Rule of Law for the Judiciary (2023), pp. 43-44, available at https://unesdoc.unesco.org/ark:/48223/pf0000387331 (last accessed on February 20, 2026); Council of ASEAN Chief Justices, Governance Framework on the Use of Artificial Intelligence (AI) for the ASEAN Judiciaries, p. 29, available at https://drive.google.com/file/d/1JLBEl52m9qi71FzZp07pF5Vw0q64ShZX/view?usp=share_link (last accessed on February 20, 2026).

^[13] Independent High-Level Expert Group on AI set up by the European Commission (2019), Ethics Guidelines for Trustworthy AI, p. 16, available at https://www.europarl.europa.eu/cmsdata/196377/AI%20HLEG_Ethics%20Guidelines%20for%20Trustworthy%20AI.pdf (last accessed on February 20, 2026); Council of ASEAN Chief Justices, Governance Framework on the Use of Artificial Intelligence for the ASEAN Judiciaries, p. 29, available at https://drive.google.com/file/d/1JLBEl52m9qi71FzZp07pF5Vw0q64ShZX/view?usp=share_link (last accessed on February 20, 2026).

^[14] Independent High-Level Expert Group on AI set up by the European Commission (2019), Ethics Guidelines for Trustworthy AI, p. 16, available at https://www.europarl.europa.eu/cmsdata/196377/AI%20HLEG_Ethics%20Guidelines%20for%20Trustworthy%20AI.pdf (last accessed on February 20, 2026); Council of ASEAN Chief Justices, Governance Framework on the Use of Artificial Intelligence (AI) for the ASEAN Judiciaries, p. 29, available at https://drive.google.com/file/d/1JLBE152m9qi71FzZp07pF5Vw0q64ShZX/view?usp=share_link (last accessed on February 20, 2026).

^[15] Inter-Parliamentary Union, Guidelines for AI in Parliaments – Ethical Principles, available at https://www.ipu.org/ai-guidelines/ethical-principles-human-autonomy-and-oversight (last accessed on February 20, 2026).

^[16] Council of ASEAN Chief Justices, Governance Framework on the Use of Artificial Intelligence (AI) for the ASEAN Judiciaries, p. 28, available at https://drive.google.com/file/d/1JLBEl52m9qi71FzZp07pF5Vw0q64ShZX/view?usp=share_link (last accessed on February 20, 2026).

^[17] Id.

^[18] Id.

^[19] Id.

^[20] UNESCO, Guidelines for the Use of AI Systems in Courts and Tribunals, p. 22, available at https://unesdoc.unesco.org/ark:/48223/pf0000396582 (last accessed on February 20, 2026).

^[21] Illustrative examples:

The Judicial Records Office may help with dataset identification for AI projects needing statistics on the types and volume of petitions being filed before the Court.

The Office of the Court Administrator may help with dataset identification for AI projects needing statistics on the Court’s personnel.

^[22] See United Nations Institute for Disarmament Research, Exploring Synthetic Data for Artificial Intelligence and Autonomous Systems: A Primer (2023).

^[23] Illustrative examples:

The SC Committee on Gender Responsiveness in the Judiciary may help with data labeling for AI projects for improving court personnel’s gender sensitivity trainings.

The Judicial Committee on Sustainability and Environmental Concerns may help with raw data cleaning and data grouping for AI projects aimed at upgrading the SC Baguio Compound’s sustainability.

^[24] UNESCO, Guidelines for the Use of AI Systems in Courts and Tribunals, p. 21, available at https://unesdoc.unesco.org/ark:/48223/pf0000396582 (last accessed on February 20, 2026).

^[25] Illustrative examples:

The SC Committee on Gender Responsiveness in the Judiciary may help with introducing positive bias for an AI project for promoting the welfare of court personnel who are single parents.

The OCA may help with eliminating biases for an AI project for evaluating personnel promotion eligibility.

^[26] Illustrative examples:

The SC MISO, in cooperation with OCA, may conduct regular testing of an AI system that has been deployed for lower courts to use in connection with their decongestion efforts.

The SC MISO, in cooperation with the SC Committee on Access to Justice in Underserved Areas, may conduct regular testing of a dialect-conversant AI chatbot that has been deployed to improve the public’s immediate access to legal aid.

^[27] Choice and choice-making here refer to AI’s reasoning and decision-making: “Reasoning and Decision Making. At the core of an AI system lies its reasoning module, which takes as input the data coming from the sensors and outputs an action to take, given the goal to achieve. This means that the data collected by the sensors need to be transformed into information that the reasoning module can understand. To continue in our example of a cleaning AI system, the camera will provide a picture of the floor to the reasoning module, and this module needs to decide whether to clean the floor or not (that is, what the best action is to achieve the desired goal). While it may seem easy for us humans to go from a picture of a floor to the decision of whether it needs to be cleaned, this is not so easy for a machine, because a picture is just a sequence of 0s and 1s. The reasoning module therefore has to:

“1. Interpret the picture to decide if the floor is clean or not. In general, this means being able to transform data into information and to model such information in a succinct way, which however should include all relevant pieces of data (in this case, whether the floor is clean or not).
“2. Reason on this knowledge in order to understand what the best action is. In this example, if the information derived from the picture is that the floor is dirty, the best action is to activate the cleaning, otherwise the best action is to stay still.” (See European Commission High-Level Expert Group on Artificial Intelligence, A Definition of AI: Capabilities and Scientific Disciplines (2018), p. 3, available at https://ec.europa.eu/futurium/en/system/files/ged/ai_hleg_definition_of_ai_18_december_1.pdf (last accessed on February 20, 2026).

^[28] Illustrative example:

The Court may engage a third-party auditor to conduct regular algorithmic audits of complex AI systems deployed for the Judiciary’s use.

^[29] Illustrative example:

The SC Committee on Computerization and Library, in coordination with the SC MISO, may conduct orientations for judges regarding the transparency levels that first- and second-level courts need to observe in relation to AI systems that they use for decongestion efforts.

^[30] Illustrative example:

The SC Committee on Computerization and Library, in coordination with the SC MISO, may act on risk reports by engaging the services of a third-party auditor and, after conducting algorithm testing, (1) regroup data, (2) relabel groups, and (3) update the models of a deployed AI system, based on the results yielded by the testing.

^[31] UNESCO, Guidelines for the Use of AI Systems in Courts and Tribunals, p. 23, available at https://unesdoc.unesco.org/ark:/48223/pf0000396582 (last accessed on February 20, 2026).

^[32] Id. at 24.

^[33] Id. at 24.

^[34] Such as the CALESA DIGITAL Project for Cooperation in Legal, Social, Academic, and Digital Advancement in the Philippines.

Document Information