PH Supreme Court Justice Leonen: AI may assist courts but never decide cases

Artificial intelligence (AI) can help improve the efficiency of the Philippine judiciary, but it should never replace judges in deciding cases, Supreme Court Senior Associate Justice Marvic Leonen said.

Speaking before judges, court personnel, and legal researchers during a July 1 briefing on "Digitalization & Artificial Intelligence in the Courts of Justice," Leonen stressed that AI should remain a tool to support judicial work—not perform the core function of adjudication.

"AI has a place in the courts: beside the judge to assist, but never on the bench to decide cases," Leonen said, according to the Supreme Court.

At present, the only AI application authorized for use in the judiciary is Scriptix, an AI-powered voice-to-text transcription tool. Leonen said any other AI application would require approval from the Supreme Court En Banc.

The briefing was part of the judiciary's efforts to familiarize court officials with the Supreme Court's Governance Framework on the Use of Human-Centered Augmented Intelligence in the Judiciary, which was approved by the SC En Banc on Feb. 18, 2026. The framework supports the Strategic Plan for Judicial Innovations 2022-2027, which aims to modernize court operations through digital technology.

Leonen, who headed the working group that drafted the framework, said AI can assist in legal research, scheduling, drafting documents, and data analytics but should never determine the outcome of cases.

"A machine may assist in research, in scheduling, drafting, and in analytics, but it may never adjudicate," he said.

He warned that while AI systems are capable of producing polished legal writing and mimicking judicial reasoning, they lack genuine understanding of the law and the human circumstances underlying legal disputes.

"The fabrication we politely call hallucinations are at once plausible but false. So a recommendation in sentencing or in a bail substitutes prediction for judgment and dresses the bias of the past in the robe of neutral science. The machine cannot reach the suffering behind the case filed, the credibility written in a witness' silence, or the justice that a mere technicality would defeat."

Leonen also distinguished digitalization from artificial intelligence, saying that transitioning from paper-based court processes to digital systems is a separate effort that can improve access to justice by reducing delays and administrative burdens.

He said technologies such as remote filing, online hearings, and digital records can make court services more accessible, particularly for litigants living far from courthouses.

"Remote filing, online hearing, digital records remove the barrier of distance so that a marginalized litigant may participate without ruinous travel," he said.

However, Leonen cautioned that digital transformation also presents challenges, including unreliable internet access and the risk of cyberattacks.

"The question is never whether to modernize. It is whether we remain sovereign over the rules that we allow," he said.

The senior associate justice also explained that AI systems rely on inductive reasoning, using historical data and statistical patterns to predict outcomes. While useful in some contexts, he said such systems are fundamentally limited because past data cannot guarantee future results and may reflect existing biases.

"It tells us to stop thinking, and to stop thinking is the one thing a court must never do. The number arrives wearing the robe of neutrality," Leonen said.

He added that reliance on historical data is particularly problematic in the justice system, where courts seek not only to apply the law but also to recognize rehabilitation and the possibility of change.

"Statistics treats the past as the best predictor of the future. That is its methodological commitment. But in law, the past may precisely be what we are trying to overcome. The entire apparatus of rehabilitation, probation, juvenile justice, restorative processes, rests on the premise that the person is not identical to what they have done."

Leonen said these concerns informed the Supreme Court's AI Governance Framework, which is anchored on six principles: human primacy, explicability, auditability, equal access, institutional sovereignty, and continuous learning.

As an example of human oversight, he cited Scriptix, emphasizing that AI-generated transcripts continue to be reviewed and verified by court stenographers before they become part of the official record.

He concluded by reiterating that while digitalization enables AI adoption, artificial intelligence remains only "a machine for prediction" whose role must always remain subordinate to human judgment in the administration of justice.