Algorithmic Accountability: Reconciling the Use of AI in Legal Pleadings with the Lawyer’s Duty of Candor under the Code of Professional Responsibility and Accountability (CPRA)

By Ijssel Dumancas and Japhet Patriz Antonio, JD1

INTRODUCTION

People often imagine a passionate attorney delivering a dramatic, impromptu monologue that completely alters the course of a trial. This births the common misconception that the courtroom is the theater where legal battles are won or lost. In reality, the overwhelming majority of legal disputes are decided long before anyone steps up to the podium. They are won on paper.

Pleadings are essentially the legal battlefield, where a lawyer outlines the facts, identifies the legal issues, and establishes the boundaries of what the court will ultimately decide. If a point is not raised in the pleadings, it generally cannot be argued later at trial. It is therefore necessary that pleadings are not dry and robotic recitations of statutes. At their core, the best pleadings tell a story. Humans are wired to respond to narratives. A lawyer with exceptional writing skills can take a chaotic set of facts and organize them into a compelling and coherent narrative that highlights the injustice his/her client has suffered.

As legal research shifts from traditional, physical books to digital resources, the process of drafting pleadings and other legal documents is naturally being transformed by this digital evolution. Now, it does not end with just research but with the very drafting of the pleadings. As the Supreme Court correctly put it, breakthroughs in artificial intelligence (AI) are affecting nearly all aspects of life and society. Such is true even in the legal profession.

AI refers to computer systems engineered to simulate human cognitive functions. These systems do not think in a human sense, but instead, they operate on advanced probabilistic logic, predicting the most textually and contextually appropriate sequence of words based on the data they were trained on. This allows them to draft essays, summarize volumes of information, write code, and mimic human conversation with fluency. However, despite its power, AI has a fundamental limitation called hallucination, wherein it confidently generates false information, fabricated data, or non-existent references while presenting them as facts because it operates on probability rather than absolute truth.

This intersection of AI and legal pleadings not only comes with advantageous practicality but also with questionable ethical resilience. In 2023, as a response, 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.

This research explores the critical relationship between AI-assisted drafting and the lawyer’s Duty of Candor as mandated under the Code of Professional Responsibility and Accountability (CPRA). It further examines how the Judiciary adopted the said frameworks through the issuance of resolutions, which are now applied in the Philippine Courts. This research aims to define and set thresholds for the ethical use of AI-generated content.

OBJECTIVES OF THE STUDY

This legal research particularly aims the following:

• To evaluate existing Philippine Supreme Court resolutions as regards emerging global trends in order to find regulatory loopholes in AI use.
• To establish a clear boundary determining where automated assistance ends and where indispensable human legal analysis must begin.
• To study and evaluate how foreign bar associations and international jurisdictions regulate generative AI to pinpoint gaps in current Philippine judicial interpretations regarding digital accountability.
• To propose actionable solutions, including mandatory disclosure rules for AI-assisted filings, to protect the integrity of the Philippine legal system.

The general goal of this study is to evaluate the ethical impact of integrating generative AI into the legal practice by analyzing how AI-assisted drafting intersects with a lawyer’s non-delegable Duty of Candor under the Code of Professional Responsibility and Accountability (CPRA), specifically assessing whether a failure to rigorously verify AI-generated content constitutes professional negligence or a breach of ethical obligations to the court, the client, and the public.

METHODOLOGY

To effectively address the tension between technological innovation and professional ethics, this study adopts a qualitative legal research design. The methodology is structured to systematically examine the compatibility of Generative AI with the stringent ethical mandates imposed on legal practitioners. Because the integration of AI into legal drafting threatens to blur the line between mechanical automation and substantive legal practice, a purely empirical approach would be insufficient. Instead, the research requires a critical interpretation of established legal norms, professional codes, and evolving global standards. By grounding the study in existing legal frameworks while simultaneously looking outward at international developments, this methodology ensures a well-rounded, evidence-based assessment of a lawyer’s non-delegable duties in the digital age.

The methodology begins with an evaluation of existing Supreme Court resolutions adopted from international organizations’ frameworks in response to lawyers’ increasing use of AI in their profession. Such analysis is geared toward determining the limitations of legal professionals’ dependence on AI and identifying which part of the court procedure the specific limitation applies to, and vice versa. The provisions of the resolutions are then compared to the latest version of the CPRA.

In addition, the methodology integrates a comparison to that of foreign courts as well as foreign ethical codes for lawyers on how other jurisdictions regulate the use of AI as a precaution against the possible erosion of ethical standards among legal practitioners. The research is expected to highlight the gaps in the Philippines’ digital accountability measures when compared to the best practices in court management and legal ethics regarding automation of other national legal systems.

Ultimately, this dual-method approach provides the empirical and theoretical foundation necessary to propose concrete legislative or judicial reforms, such as mandatory disclosure rules for AI-assisted filings.

DISCUSSION

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”. As of the present, A.M. No. 25-11-28-SC Governance Framework on the Use of Human-Centered Augmented Intelligence in the Philippine Judiciary, which is a Supreme Court En Banc Resolution, is the framework that provides a comprehensive regulation on the integration of Artificial Intelligence. It has three main pillars: fairness, accountability, and transparency, strictly emphasizing that AI must serve as a support tool rather than a replacement for human judgment.

The memorandum gives five specific regulations and procedural rules that directly apply to lawyers and legal practitioners. First, members of the bar are mandated to disclose the extent of their use of AI in drafting their pleadings, motions, or other legal submissions. The disclosure shall include the specific AI tool and version used, the extent or degree of their reliance on the AI’s output, the level of human control, review, and oversight that was exercised over the final output, a confirmation that the unaltered AI output for potential judicial inquiry or audit is preserved, a formal statement affirming full compliance with the Supreme Court’s AI framework, and that lawyers assume full responsibility for the content. Second, an absolute professional and personal accountability is placed squarely on the shoulders of the signing counsel. Guided by the principle of human paramountcy, practitioners are required to act as the primary human-in-the-loop, and are expected to have verified every factual assertion and legal authority before it is submitted to a tribunal. Third, the framework prohibits lawyers from uploading, feeding, or processing any confidential client information, privileged communications, or sensitive personal data into public or unsecured cloud-based AI platforms. This must be consistent with existing data privacy laws and an adherence to the upholding of client-lawyer confidentiality. Lastly, practitioners are strictly barred from using generative AI to produce complete legal arguments without rigorous human intervention, and any attempt to use AI to generate deepfakes or manipulated evidence to mislead the court will be treated as severe professional misconduct.

In the United States of America, as well as in Canada, many Federal District Courts have implemented strict standing orders requiring lawyers to file an “AI Certification.” This forces lawyers to state whether generative AI was used to draft a filing and certify that a human verified the accuracy of all legal citations. This is somewhat similar to the resolution issued by the Supreme Court. In the United Kingdom (U.K.), on the other hand, disclosure is only mandated if a judge explicitly probes a suspicious filing or if a hallucination is caught. Despite the lack of upfront disclosure as to the provisions of the U.K. Courts and Tribunals Judiciary’s Artificial Intelligence (AI) Guidance for Judicial Office Holders, the U.K. focuses heavily on the obligation not to mislead the tribunal. This was manifested in the landmark case of R (Munir) v. Secretary of State.

When comparing the ethical codes and formal guidance governing lawyers’ use of AI, jurisdictions globally utilize remarkably similar baseline ethical duties of competence, confidentiality, candor, and supervision. However, they diverge in how proactively transparent lawyers must be. While the Philippines adopts a highly bureaucratic, preventative approach, Western jurisdictions largely depend on existing professional ethics codes to punish errors ex-post facto.

In the Philippines, the CPRA is a broad ethical code that covers a vast array of how lawyers should conduct themselves, which range from financial dealings and social media decorum to courtroom behavior. As the code uses a general tone, the obligations of law practitioners set in A.M. No. 25-11-28-SC are not explicitly detailed. Conversely, a significant limitation of the AI framework is that it lacks its own independent disciplinary or penal mechanism to punish erring lawyers. It is limited to identifying technological infractions and data breaches, and it must ultimately refer those violations back to the CPRA to be penalized through traditional bar discipline, suspension, or disbarment.

CONCLUSION AND RECOMMENDATIONS

The Philippine AI framework, codified under OCA Circular No. 47-2026, positions itself as one of the most centralized and preventative frameworks in the world. Rather than regulating the technology from a market or consumer standpoint, the Philippine Supreme Court views AI primarily through the lens of institutional risk and judicial integrity. For practitioners, this translates into heavy, non-negotiable administrative duties requiring them to file detailed, plain-language disclosures alongside any AI-assisted submissions.

By blending international best practices with the ethical bedrock of the Code of Professional Responsibility and Accountability (CPRA), the Philippine judiciary can transition this landmark framework from a static policy into a highly enforceable system that aligns with the CPRA’s Duty of Candor.

• Require a human validation protocol. While A.M. No. 25-11-28-SC and the CPRA’s Duty of Candor hold a lawyer strictly liable for AI hallucinations, neither text explicitly defines what legally constitutes sufficient human verification. Without clear boundaries, honest mistakes could be treated as intentional acts of deception. To fulfill the Duty of Candor, lawyers must certify that they have manually verified the citations using a Supreme Court-approved digital law library. This adopts the proactive stance that legal citations have been verified by a human eye.
• Standardize disclosure through electronic-filing certifications. The judiciary should automate candor by integrating a mandatory selection of machine-readable categories of how they used AI before a pleading can be uploaded. This is to prevent lawyers from evading the framework’s requirement to disclose AI use in plain language by reason of ambiguity.
• Mandate a “Chain of Logic” log. A.M. No. 25-11-28-SC orders lawyers to preserve raw AI outputs for potential judicial audits. However, raw copy-pasted text logs are incredibly easy to format, edit, or selectively delete, making them practically useless as forensic evidence if a lawyer is accused of a breach of candor. Rather than forcing lawyers to manually save and track their AI text files, the Supreme Court should require them to use professional legal software that automatically creates a secure and permanent history of their work.
• Issue a list of allowable software enterprises. The Supreme Court’s permanent AI Committee should publish and regularly update an official Judicial list of AI software. By negotiating institutional data privacy riders with tech vendors, the Court can ensure local practitioners use software with strict closed-loop architectures. This guarantees that any inputted Philippine case files are instantly deleted or sandboxed, allowing lawyers to achieve maximum efficiency without accidentally compromising client-lawyer privilege.
• Include a penalty provision. While A.M. No. 25-11-28-SC forbids the use of AI tools for behavioral manipulation, it lacks explicit definitions and procedural workflows for handling a lawyer who deliberately uses AI to deceive the bench. The Supreme Court should clearly define deceptive AI practices under the umbrella of Prohibited AI Systems in Part I of the framework.

The transition of the Philippine legal system into the digital era is anchored by two foundational frameworks issued by the Supreme Court: the Code of Professional Responsibility and Accountability (CPRA) and the Governance Framework on the Use of Human-Centered Augmented Intelligence in the Judiciary (A.M. No. 25-11-28-SC). The connection between these two frameworks is highly collaborative. The CPRA sets the baseline moral and professional standard for lawyers, while the AI Framework brings those rules into the digital age by dictating exactly how core values like competence, confidentiality, and truthfulness apply to software and algorithms. Technology should change how we practice law, but it must never change why we practice law. AI is an exceptional co-pilot for handling the administrative and repetitive burdens of litigation, but the moral reasoning, ethical judgment, and ultimate accountability must remain exclusively in human hands.

BIBLIOGRAPHY

Statutes

The 1987 Constitution of the Republic of the Philippines, Article VIII, Section 5(5).

Case Laws

R (Munir) v. Secretary of State for the Home Department, UKUT 81 (IAC), 2026.

Administrative Regulations

Courts and Tribunals Judiciary, Artificial Intelligence (AI): Guidance for Judicial Office Holders, 2025.

Northern District of Texas Civil Rules, N.D. Tex. Civ. R. 7.2(f).

The Supreme Court of the Philippines, A.M. No. 22-09-01-SC Code of Professional Responsibility and Accountability, 2023.

The Supreme Court of the Philippines, A.M. No. 25-11-28-SC Re: Proposed Governance Framework on the Use of Human-centered Augmented Intelligence in the Judiciary, 2026.

Journal Articles

American Bar Association Standing Committee on Ethics and Professional Responsibility, Formal Opinion 512, 2024.

Online Sources

Vermont Law Review “Mandatory AI Disclosures: Enforcing A Uniform Standard” https://lawreview.vermontlaw.edu/mandatory-ai-disclosures-enforcing-a-uniform-standard/#:~:text=%5B1%5D%20Judge%20Brantley%20Starr%20of,with%20traditional%20legal%20research%20tools. (accessed May 2026).