The Commission on Higher Education’s (CHEd) Memorandum Order (CMO) No. 21, series of 2026, gives Philippine colleges and universities a framework for using artificial intelligence (AI). Its strongest feature is the room it leaves for each higher education institution (HEI) to develop policies suited to its programs, needs, resources, and readiness. A single rule cannot serve every campus equally well. But institutional autonomy must lead to clear guidance, not a patchwork of vague prohibitions.
The order’s human-centered principle offers a sound starting point: AI should strengthen human capabilities while protecting rights and well-being. Faculty members therefore need to teach students how to use AI responsibly-for example, to test ideas, improve understanding, and question an output’s accuracy-without surrendering their own judgment. This matters beyond graduation, as students enter workplaces where AI use will increasingly be expected.
Furthermore, the CMO reiterates that any HEI deploying high-risk AI systems must establish an Institutional AI Governance Framework outlining clear internal approval pathways and oversight. There must also be a requirement for human oversight in all final decisions, and no AI system may autonomously determine academic outcomes without meaningful human review. Documentation of system logic and purpose, bias and impact assessments, and periodic review must likewise be carried out to ensure continued fairness and reliability.
Transparency is equally necessary. Students should disclose AI use in academic submissions, with HEIs deciding how that disclosure works. A brief statement explaining which tool was used and for what purpose can make expectations understandable. Yet disclosure is not a transfer of authorship. Students and researchers remain responsible for their arguments, sources, data, and conclusions.
I am especially concerned about policies that turn an AI detection score into a verdict. No detector is 100 percent reliable. Research has found that detectors can falsely classify writing by non-native English speakers as AI-generated; other evaluations show that AI-generated text can evade detection. Tertiary educators, including graduate school professors, should not impose very low percentage cutoffs or treat a score alone as proof of misconduct. That would risk penalizing honest students while offering false confidence about academic integrity. As educators, we have our own creative ways of determining how students write, create, and articulate their thoughts.
HEIs can put the policy into practice in four ways. First, publish course-level examples of permitted, restricted, and prohibited AI use, alongside a simple disclosure format. Second, train faculty and students to check AI outputs for invented references, bias and errors, and to protect personal and research data. Third, assess learning through drafts, research logs, oral defenses, and other evidence of a student’s reasoning. Fourth, require meaningful human review and an opportunity to respond before any academic penalty or consequential decision involving AI, including grading or admissions.
CHEd has set a useful national direction. Its success will depend on whether universities make responsible AI use teachable, fair, and genuinely accountable to people.