Education & Universities

AI policy for universities in Malaysia: what to include

What a workable AI policy for a Malaysian university contains: scope, permitted uses, declaration, assessment rules, PDPA and review, with a drafting process.

  • By
  • Updated
  • 6 min read
  • Facts checked
Short answer

A workable university AI policy defines which AI tools and users it covers, states permitted and forbidden uses, requires students to declare AI use, sets assessment rules, protects personal data under the PDPA, trains staff and is reviewed every term or year. Keep it short enough that students read it.

Part of the guide: AI for Teachers and Lecturers in Malaysia: Use, Assess and Set Rules

On this page
  1. What Core Components Must a University AI Policy Address?
  2. How Does the Policy Address Academic Integrity and Assessment?
  3. What Specific Steps Should Be Taken to Draft the Policy?
  4. How Must the Policy Ensure Data Privacy Under Malaysian Law?
  5. What Role Does Faculty Training Play in Policy Success?
  6. How Should the Policy Be Reviewed and Enforced?
  7. Where to go next
  8. Sources
  9. Frequently asked questions

A university AI policy that students and lecturers can follow is short, specific and reviewed often. This guide lists what it should contain and how to draft it in six steps. Read the primary documents yourself: the Ministry of Higher Education issued guidelines on generative AI in 2024, and the Malaysian Qualifications Agency Advisory Note 2/2023 recommends prudent and responsible use.

What Core Components Must a University AI Policy Address?

A comprehensive AI policy cannot be a single document; it must be a framework addressing various operational facets of the university. Key components ensure that AI use is managed systematically across teaching, research, and administration.

Scope Definition: The policy must clearly define what "AI" encompasses. This includes generative AI tools (e.g., ChatGPT, Copilot), analytical AI, machine learning models, and any software that automates cognitive tasks. The scope must cover all users: students, faculty, and administrative staff.

Permitted Uses: This section outlines where and how AI is allowed. For instance, AI might be permitted for brainstorming, summarizing complex texts, or generating initial code drafts in technical subjects. Conversely, it must explicitly state where AI use constitutes academic misconduct (e.g., submitting AI-generated essays as original work without attribution).

Declaration of AI Use: A mandatory declaration mechanism is necessary. Students and researchers must be required to disclose when and how AI tools were used in their submissions or research processes. This moves the policy from prohibition to transparency.

Assessment Rules: This is perhaps the most critical area for academic integrity. The policy must dictate how assessments are designed to mitigate AI reliance. This includes shifting assessment methods towards oral exams, in-class problem-solving, or assignments requiring deep, context-specific local knowledge that current AI models struggle to replicate authentically.

PDPA Data Privacy Compliance: Given the sensitivity of student and staff data, the policy must strictly adhere to the Personal Data Protection Act (PDPA) of Malaysia. Any AI system deployed by the university must guarantee that personal identifiable information (PII) is not used to train external, public AI models without explicit consent.

Staff Training and Capacity Building: A policy is ineffective if staff do not understand it. Mandatory, ongoing training is required to equip faculty and staff with the skills to identify AI-assisted work, design AI-resistant assessments, and use AI ethically in their own workflows.

Review and Enforcement Cycle: The policy must not be static. It requires a defined review cycle (e.g., every 18–24 months) to adapt to rapidly evolving AI technology. Enforcement mechanisms must be clearly defined, linking policy breaches to existing disciplinary procedures.

How Does the Policy Address Academic Integrity and Assessment?

The primary concern for any university is maintaining the value of the degree and the integrity of the learning process. The policy must move beyond simply banning tools and instead focus on redefining learning outcomes in the age of AI.

The policy should establish clear boundaries between acceptable augmentation and unacceptable substitution.

AI Use Category Permissible Use Example Non-Permissible Use Example Required Declaration
Drafting/Writing Generating outlines or refining grammar. Submitting an entire essay generated by AI. Yes (Specify prompt usage)
Coding/Programming Debugging small sections of code. Generating a complete, complex software module. Yes (Specify tool and function)
Research/Literature Summarizing abstracts from large datasets. Presenting AI-generated synthesis as original critical review. Yes (Cite AI as a research assistant)
Assessment Using AI to check for plagiarism against known databases. Using AI to grade subjective assignments without human oversight. N/A (Administrative function)

The policy must clearly state that the ultimate responsibility for the work remains with the student or researcher, regardless of the tools employed.

What Specific Steps Should Be Taken to Draft the Policy?

Drafting such a policy requires a multi-stakeholder, iterative approach rather than a top-down mandate. A phased implementation is recommended.

  1. Establish a Cross-Functional Steering Committee: This committee must include representatives from Academic Affairs, IT/Data Security, Legal Counsel, Student Affairs, and Faculty Senate.
  2. Review Existing Framework: Thoroughly review the guidelines issued by the Ministry of Higher Education and the recommendations within MQA Advisory Note 2/2023 to ensure compliance with current national directives.
  3. Conduct Stakeholder Consultations: Hold targeted workshops for different user groups (undergraduate, postgraduate, research staff) to gather practical concerns and suggested use cases.
  4. Draft Core Principles: Develop the high-level ethical guidelines (e.g., transparency, accountability, fairness) before detailing specific rules.
  5. Develop Implementation Guidelines: Create supplementary documents detailing how faculty should modify assignments and how IT should monitor usage, linking these to the main policy.
  6. Pilot and Refine: Implement the draft policy in a controlled pilot program (e.g., one specific department or course level) before a full university rollout.
  7. Communicate and Train: Launch the policy with mandatory, accessible training modules for all staff and students.

How Must the Policy Ensure Data Privacy Under Malaysian Law?

Adherence to the PDPA is non-negotiable. The policy must treat AI interactions as data processing activities.

The policy must mandate the following regarding data handling:

  1. Data Minimization: Only collect and process the minimum amount of data necessary for the stated educational or research purpose.
  2. Consent Mechanisms: For any use of student data in an AI system that is not strictly necessary for core academic delivery, explicit, informed consent must be obtained.
  3. Vendor Vetting: If the university uses third-party AI vendors, the policy must require rigorous due diligence to ensure the vendor’s data processing agreements comply fully with Malaysian data residency and privacy standards.
  4. Anonymization Protocols: Where possible, all data used for training or analysis must be rigorously anonymized before entering any AI platform.

What Role Does Faculty Training Play in Policy Success?

Faculty are the frontline implementers of the policy. If they are unsure how to apply the rules, the policy will fail or be circumvented. Training must be practical, not theoretical.

Effective training should cover:

  1. AI Literacy: Understanding the capabilities and limitations of current AI models (e.g., hallucination risk, bias).
  2. Assessment Redesign: Practical examples of transforming traditional take-home essays into AI-resistant assignments (e.g., requiring local Malaysian case studies, oral defenses).
  3. Detection Tools: Awareness of available detection software, while acknowledging that these tools are imperfect and should not be the sole basis for disciplinary action.
  4. Ethical Prompt Engineering: Teaching staff how to guide students to use AI as a tool for learning rather than a shortcut to completion.

How Should the Policy Be Reviewed and Enforced?

A policy that is never updated is obsolete. The review cycle must be proactive, not reactive.

The policy should establish a clear enforcement hierarchy:

  1. Level 1 (Minor Infraction): First-time, low-impact misuse (e.g., failing to cite AI use in a minor section). Action: Mandatory retraining and formal warning.
  2. Level 2 (Moderate Infraction): Substantial reliance on AI for core components of an assignment without disclosure. Action: Grade reduction or resubmission requirement.
  3. Level 3 (Severe Infraction): Presenting entirely AI-generated work as original research or thesis submission. Action: Disciplinary action according to university statutes.

The review cycle should be triggered by:

  • Major advancements in AI technology.
  • New guidance issued by the Ministry of Higher Education or MQA.
  • Significant incidents of policy breach.

Where to go next

Lecturers who need to apply the policy in class can start with AI for teachers and lecturers and our AI training for universities. For assessment rules, see assessing students when everyone uses AI. Personal data handling is covered in PDPA and AI integration. Private institutions registered with HRD Corp may be able to apply for staff training under HRD Corp claimable courses, subject to approval.

Sources

Links below checked November 2026.

Frequently asked questions

Who should write a university AI policy?

A small group of academic leaders, lecturers, IT and the data protection contact, with student input. One owner keeps it current.

Should a university ban generative AI?

Bans are hard to enforce. Most policies set permitted uses, require declaration and change how work is assessed. Follow your own institution and regulator guidance.

How often should the policy be reviewed?

At least once a year, and whenever the tools or the regulator guidance change.