Are our schools ready for AI?
Artificial intelligence has embedded itself into the daily rhythms of Philippine basic education, with students using conversational chatbots to structure essays and teachers relying on automated platforms to generate quizzes. For a few years since ChatGPT’s release in 2022, the Department of Education operated without a unified stance, leaving individual schools to decide whether automated tools should be embraced or restricted. DepEd Order No. 003, series of 2026, entitled “Foundational Guidelines on Artificial Intelligence (AI) in Basic Education,” attempts to resolve the confusion by introducing a comprehensive national regulatory framework governing artificial intelligence. The policy mandates a central ledger of approved technologies, risk categorizations, and privacy reviews before software reaches students.
The framework represents a significant administrative milestone. It may be too early to tell whether written mandates can be operationalized through proper governance and are able to resolve classroom challenges. Institutional readiness rests entirely on whether the state can establish an active operational lifecycle that defines risk before use, detects harm during use, and corrects harm after use.
Mitigating AI risk before it’s too late
A proactive approach to technology governance requires evaluating potential dangers well before tools enter the classroom. DepEd addresses the challenge by separating benign educational software from high-stakes systems that directly influence student grades, disciplinary records, or scholarship opportunities. Mandating a Privacy Impact Assessment for new applications ensures that student data remains secure from commercial exploitation, forcing school administrators to deliberate on data collection practices before introducing a platform to vulnerable young learners. These pre-use screenings offer structural protection, yet the efficacy of a centralized database is limited because a centralized ledger can only catalog software officially deployed by the school. It does not possess a mechanism to track personal AI models that students access on private smartphones or home networks. When a learner uses an unlisted, consumer-grade chatbot at home to ghostwrite an assignment, the registry remains completely unaware. Abstract risk classifications look clean in a central office, but consumer technology easily breaches the administrative perimeter, which renders pre-deployment filters secondary to real-time classroom realities.
Academic dishonesty
Live instructional interactions reveal that digital maturity assessments and self-reporting mechanisms are insufficient instruments for verifying classroom integrity. Evaluating readiness based on infrastructure quality or institutional leadership does not give an educator the ability to decipher whether a submitted paragraph reflects a student’s authentic voice or a machine’s polished output. Relying heavily on honest disclosures creates an honor system that can be easily manipulated by users who report assistance selectively or omit disclosures entirely. The burden of monitoring active learning environments falls on classroom teachers, transforming educators into accidental technical auditors. Expecting a teacher to cross-reference every sentence with automated detection software is unsustainable within an educational system burdened by severe human resource shortages. The issue extends far beyond academic dishonesty to actual cognitive and pedagogical harms. For instance, an automated essay-grading platform can generate shallow, culturally detached feedback that misguides a learner if left unmonitored. When an unverified tool misinterprets student writing or provides flawed explanations, the teacher remains the only barrier protecting the student from learning degradation.
AI and false flags
The true vulnerability of automated systems becomes undeniable when a student suffers an unfair academic setback and lacks a clear mechanism for redress. For instance, an AI-driven proctoring tool might falsely flag a student from a rural province for suspicious movement due to an unstable internet connection or a noisy household. If the school accepts the automated judgment without a rigorous appeal process, the student faces undeserved disciplinary action or a ruined grade. Resolving the dilemma requires immediate human intervention that overrides the machine’s verdict and protects the learner’s academic standing. Effective post-harm correction requires an explicit escalation protocol that holds technology providers and school officials accountable for systemic failures. When an automated grading tool produces flawed outcomes, the affected student must have the right to demand an independent human review by a panel of educators. School heads must possess the authority to pause the deployment of any platform under investigation, ensuring that questionable software is suspended while engineers resolve the underlying technical bias. True correction demands that schools update the national registry, rectify erroneous academic records, and permanently retire any software that repeatedly fails to meet ethical learning standards.
Where DepEd comes in
Moving to a functional safety network requires investing in concrete classroom routines rather than general awareness seminars. DepEd must refine its instructions to help teachers distinguish between prohibited automation, acceptable collaborative assistance, and instances where students are expected to critically evaluate AI outputs. Standardizing these rules across different subjects allows educators to manage technology without needing to reinvent procedures for every assignment. Furthermore, compensating for the limitations of automated detection requires a profound shift in how teachers design student assessments. Future examinations must prioritize formats that software cannot easily replicate, emphasizing oral defenses, in-class writing sessions, and localized case studies that require real-world engagement. The objective focuses on designing learning experiences where human comprehension becomes visible and undeniable rather than winning a technical arms race against generative tools.
Another essential element of an operational strategy involves establishing a clear three-tiered harm escalation pathway across the education bureaucracy. Minor disclosure omissions should be resolved directly between the teacher and the student through standard classroom counseling. Serious breaches of academic integrity or recurring algorithmic bias must be escalated to the school principal for formal administrative review. Major institutional crises, including severe data breaches or systemic proctoring failures, must be handled directly by regional division offices and the designated responsible officer. Through these concrete operational steps, DepEd can transform an aspirational set of rules into a practical framework that actively safeguards Philippine classrooms.
Josh Huesca is an AI Governance Fellow at AI Safety Diliman and a senior pursuing a BS in Health Sciences with a Minor in Management at the Ateneo de Manila University. His interests sit at the intersection of strategy, technology, and finance, with a focus on strengthening institutional responses that shape health, environmental outcomes, and public life. [LinkedIn]
The author acknowledges the supervision of Atty. Edsel Tupaz, Lead Researcher and Fellow at AI Safety Diliman, Ze Shen Chin of AI Standards Lab and Research Affiliate at the Oxford Martin School AI Governance Initiative, as well as Jericho Roland Sueno, Lenz Dagohoy and Lexley Villasis of AI Safety Diliman.
AI Safety Diliman is supported by Kairos, an AI safety field-building nonprofit focused on strengthening the global AI safety and policy talent pipeline, and is fiscally sponsored by the Berkeley Existential Risk Initiative, an entity that supports academic and field-building efforts to reduce catastrophic risks associated with advanced technologies.
