Supreme Court Senior Associate Justice Marvic M.V.F Leonen discussed in a recent lecture the Supreme Court’s (SC’s) Governance Framework on the Use of Human-Centered Augmented Intelligence in the Judiciary (AIGovernance Framework).

The framework is a comprehensive guide on the responsible use of human-centered augmented intelligence in the judicial system, based on ethical principles such as fairness, accountability, and transparency.

Leonen explained that under the AI Governance Framework, court personnel are prohibited from using AI tools unless approved by the Court.

Presently, the only approved AI tool is Scriptix, a voice-to-text transcription tool used by court stenographers.

SAJ Leonen stressed that AI differs fundamentally from human intelligence. “[AI] can follow complex procedures…but it cannot understand,” he stressed.

While acknowledging that human intelligence is difficult to define, SAJ Leonen identified several qualities that distinguish it from AI.

He noted that human intelligence is embodied, emotional, contextual, situational, and cultural—qualities that shape how people, perceive, interpret, and respond to the world.

He warned that AI’s growing capabilities present serious risks when used without caution.

One risk, he explained, is overreliance, as people increasingly delegate thinking and writing tasks to AI systems, they may gradually lose the skills they no longer practice.

Citing research on neuroplasticity, he warned that abilities that are not regularly exercised can deteriorate over time.

“What you do not use, you will lose,” he warned, he also cautioned against allowing AI to narrow human curiosity and critical inquiry.

Because AI can quickly generate precise answers, users may become less inclined to explore alternative perspectives, ask follow-up questions, or challenge their own assumptions.

SAJ Leonen said that equally concerning are the issues of hallucination and misgrounded information. AI systems may produce nonexistent cases or citations.

They may also correctly cite real sources but misinterpret their meaning.

Illustrating this risk in the legal profession, SAJ Leonen offered an example:

“The case exists, the citation is correct, but the authority it interprets does not support the proposition.”

These errors can become especially dangerous in the law, where inaccurate

interpretations may eventually be treated as authoritative.

He warned: “A justice uses AI, and says, this is the rationale. Then somebody cites that again. Then somebody cites that again. What is wrong becomes a whole doctrine.”

He likewise cautioned against excessive multitasking through AI tools, saying that constantly dividing one’s attention can weaken the ability to focus deeply and think critically.

Beyond these risks, he noted that AI systems rely heavily on statistical models that categorize individuals according to patterns and probabilities.

He cautioned that such approaches may overlook the complexity and uniqueness of human experience, emphasizing that “statistics is a science of the population, not of the individual.

”Every person, he explained, embodies a distinct combination of identities, experiences, and circumstances that may not be adequately captured by statistical classifications.

In balancing these concerns, SAJ Leonen urged participants in the lecture to go beyond simply using AI and instead cultivate what he described as “critical use,” which requires independently evaluating AI outputs and recognizing when they are wrong.

Users was also advised to employ AI strategically, such as by expanding research, crafting prompts carefully, testing assumptions, and identifying gaps in reasoning rather than simply seeking ready-made answers.

According to SAJ Leonen, AI-generated responses should be treated only as preliminary inputs that require human review and evaluation.

“AI should not provide the answer, but it’s supposed to challenge your own thinking,” he said.

He further encouraged users to maintain audit trails, reflect on their decision-making processes, continue reading and writing independently, and ensure they can perform tasks themselves rather than becoming dependent on machines.