Can Philippine Lawyers Use AI? What A.M. No. 25-11-28-SC and the CPRA Require

In short: nothing in Philippine rules prohibits a lawyer from using AI. But the Supreme Court's AI framework, A.M. No. 25-11-28-SC, expressly covers lawyers who deal with the courts. It makes the user personally responsible for what the tool produces, and it says that relying on AI can neither excuse a violation nor mitigate the penalty for it. The Code of Professional Responsibility and Accountability (CPRA) already forbids misleading a court about the existence or contents of any legal authority. In 2025 the Supreme Court began disciplinary proceedings on its own against a lawyer whose petition cited decisions that did not exist or did not say what was claimed.

What is A.M. No. 25-11-28-SC?

It is the Supreme Court's Governance Framework on the Use of Human-Centered Augmented Intelligence in the Judiciary, adopted by the Court En Banc on February 18, 2026. It sets out how AI may be used in the justice system, around three core ethical guidelines: fairness, accountability and transparency. Its starting premise is that AI assists people and does not stand in for them. In the Framework's words, it "should not replace human discernment."

Does the Framework apply to lawyers?

Yes. Its scope reaches beyond judges and court personnel to court users, whom it defines as "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."

For court users, it applies "in all their dealings and submissions with the courts or other offices in the Judiciary, including the preparation and submission of court-bound documents." A pleading drafted with the help of an AI tool falls squarely inside it.

What does it require of a lawyer who uses AI?

Four sentences in the Framework's accountability principle do most of the work:

1. You own the output. "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."
2. A human reviews everything. "Notwithstanding any automated or semi-automated process of the said tool, all outputs must be reviewed and approved by human beings."
3. AI is not a defence or a mitigating circumstance. The use of AI "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."
4. You cannot blame the tool. "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."

For judges, the Framework adds a hard limit. AI output may never be "the sole, primary or determinative basis of any adjudicatory outcome."

Do lawyers have to disclose AI use in their pleadings?

The Framework's detailed disclosure requirement applies to members of the Judiciary and court officials and employees. When they use an AI tool, they must disclose it, including which tool was used, for what purpose, to what extent, and with what human oversight. The Framework also provides that "the disclosure shall be in accordance with guidelines to be issued by the Court."

The Framework leaves the specific format for attorney filings to future Court guidelines. Pending their issuance, counsel remains fully bound by the accountability rules set forth above.

What the CPRA already requires

The Framework adds to duties that lawyers already owe under the CPRA (A.M. No. 22-09-01-SC). It does not replace them. Three provisions matter most when AI is involved:

• Candour about authority, Canon II, Section 8. "A lawyer shall not misquote, misrepresent, or mislead the court as to the existence or the contents of any document, argument, evidence, law, or other legal authority, or pass off as one's own the ideas or words of another, or assert as a fact that which has not been proven." A citation to a case that does not exist, or to a case that does not hold what it is cited for, falls within this section however the citation was produced.
• Thorough research, Canon IV, Section 1. "A lawyer shall be thorough in research, preparation, and application of the legal knowledge and skills necessary for an engagement."
• Client confidences, Canon III, Sections 27 and 28. A lawyer "shall maintain the confidences of the client, and shall respect data privacy laws." A lawyer also "shall not reveal the confidences of the client, including data from the client's files," except in the cases the Code lists, such as with the client's written informed consent. This bears directly on what you paste into any AI tool.

What happens when the citations are wrong

In Suazo v. Tri Megaforce Security Services Corporation (G.R. No. 276106, August 20, 2025, Second Division), the petition cited "what appeared at first reading to be 18 decisions of the Court." The Court checked each one. It found that they "either do not exist, contain General Register (G.R.) Numbers for other unrelated cases and erroneous promulgation dates, pertain to cases that have no relation to the issues in this case or worse, or contain doctrines that oppose rather than support" the petitioner's arguments.

The Court's assessment was direct:

"At best, it shows shoddy legal research not befitting any lawyer worth their salt, and at worst, it evinces an intent to deceive the Court into ruling in a client's favor by inventing and misrepresenting case law to support one's arguments."

The Court also noted that "misrepresenting the law is not only a punishable act under the CPRA but a violation of the Lawyer's Oath." It denied the petition, instituted on its own motion an administrative disciplinary complaint against the petitioner's counsel for a possible violation of Canon II, Section 8, and referred the matter to the Integrated Bar of the Philippines.

The decision does not say how the citations were produced. The lesson holds either way: the Court checks what lawyers cite, and a citation that fails the check is the lawyer's problem, not the source's.

A pre-filing checklist for AI-assisted work

• Confirm every authority exists. Match the G.R. number, the case title and the promulgation date against the decision itself, not against the AI's summary.
• Read the passage you rely on. A real case can still be cited for something it does not hold, or even for the opposite.
• Check what came after. Confirm that the doctrine has not been modified, abandoned or superseded by a later decision.
• Mind what you share. Do not put client confidences or data from the client's files into a tool unless it is permitted under Canon III.
• Own the final text. Review and approve every AI-assisted passage as your own work, because under the Framework it is.

What to watch next

The Framework directs the Integrated Bar of the Philippines to "monitor the legal profession's engagement with and use of AI tools, and regularly report its findings to the Supreme Court." It also encourages the IBP to use its Continuing Legal Education program to train lawyers on the ethical and responsible use of these tools. Together with the disclosure guidelines still to be issued, these are the next places where the rules for lawyers will take shape.

Questions lawyers ask

Is it ethical for a Philippine lawyer to use AI?

Yes, provided the lawyer stays responsible for the result. The Framework treats AI as an aid to human judgment, and the CPRA's duties of candour, thoroughness and confidentiality apply to AI-assisted work as they apply to any other work.

Does A.M. No. 25-11-28-SC apply to lawyers, or just to judges?

It applies to lawyers too. Its scope includes "court users," defined to include members of the Philippine Bar and Shari'ah Bar, in their dealings and submissions with the courts, including court-bound documents.

Can a lawyer say the AI made the mistake?

Not as a defence. The Framework says AI use neither excuses the user's responsibility nor serves as a ground to mitigate liability, and that a member of the Bar may not evade responsibility for a CPRA violation by claiming the AI caused it.

Has the Supreme Court acted against lawyers over non-existent citations?

Yes. In G.R. No. 276106 (August 20, 2025), the Court found that a petition cited decisions that did not exist or were misrepresented. It instituted a disciplinary complaint against counsel on its own motion and referred the matter to the IBP.

How Intellegal fits

Intellegal is built so that the checking step is fast. Every case it cites links to the Philippine decision itself, so you can open the authority and read the passage rather than take a summary's word for it. Deep Synthesis checks its own citations: a verified citation becomes a link to the source, and one it cannot match is flagged or removed rather than left in silently. Case-Law Analytics and Case Contrast let you trace a decision and line it up against later cases, to see whether the point still stands.

Intellegal supports the lawyer; it does not replace the lawyer. The responsibility the Framework describes stays with you, and Intellegal is built to make meeting it practical.

Primary sources

• A.M. No. 25-11-28-SC, Governance Framework on the Use of Human-Centered Augmented Intelligence in the Judiciary (February 18, 2026): elibrary.judiciary.gov.ph
• A.M. No. 22-09-01-SC, Code of Professional Responsibility and Accountability: sc.judiciary.gov.ph
• Suazo v. Tri Megaforce Security Services Corporation, G.R. No. 276106 (August 20, 2025): sc.judiciary.gov.ph

Try Deep Synthesis →