Image credit: Author using ChatGPT
When plagiarism is no longer the whole problem
Since the advent of generative AI, its impact on higher education has been quite profound. In a short span, it has become a research assistant, coding tutor and on-demand source of explanation for students and instructors alike. It also helped simplify complex information, streamlining repetitive tasks, and enabling personalised learning experiences. When used judiciously, it can augment intellectual work. However, the same characteristics that make generative AI useful have thrown the underlying principles of traditional assessment into disarray. Thus, the real issue for universities is not to prevent the use of AI in academia but to rethink the definition of academic integrity and assessment so that technology does not become an alternative to evidence-based learning.
Universities initially treated the issue as another academic integrity problem, something that can be curtailed with strict plagiarism checks. Generative AI can execute almost any task conveniently and imitate the visible product of intellectual work. The more vexing issue, therefore, is whether conventional assessment methods can still reveal a student’s own intellectual capacity. The assignment on the professor's desk may exhibit the outward signs of competence, but provide only a superficial indication of what the student is actually capable of. This creates an assessment validity problem where the instruments used to assess learning may or may not yield accurate results. Once this link becomes uncertain, the problem becomes endemic to the system and ratcheting up plagiarism rules does not fix it.
Pakistan’s Predicament
Students' behaviour has shifted beyond institutional policies and numbers are telling in this regard. According to a survey conducted by the UK-based Higher Education Policy Institute, 94 per cent of students reported using AI to help with assessed work (Stephenson & Armstrong,2026). From being an aberration, AI use is becoming a part of the normal academic environment. Pakistan is facing a similar predicament but from a more complicated institutional baseline. Academic integrity here is traditionally structured around accreditation rules, similarity reports and the authenticity of submitted text. Generative AI unsettles these structures. A student may produce work fluent enough that it goes beyond his own cognitive and intellectual capacity, while an AI detector may flag an original thought or produce a wrong judgement about authorship.
There is also an equity problem that Pakistani universities are unable to overlook. While access to premium content and digital literacy, and faculty competence remain uneven, technological innovation can exacerbate existing educational inequities if AI is embedded into coursework. Recent evidence in Pakistan shows the magnitude of problem. A study of 318 postgraduate students across six public and private universities in Punjab found that inadequate instructor guidance, pressure for higher grades and fear of failure were key contributors to academic misconduct. Importantly, the researchers did not observe statistically significant differences in academic integrity beliefs between students at public and private universities, suggesting that the issue may be more systemic than institutional (Ahmad et al.,2026).
Pakistan has policy, but it must travel to grassroot levels
Pakistan’s regulatory response has therefore evolved. The Higher Education Commission’s (HEC) Framework on the Use of Generative AI in Higher Education Institutes is listed as an official academic framework (Higher Education Commission, 2026). Rather than imposing a blanket ban, it calls for greater institutional oversight to ensure the acceptable use of AI, frequent training of faculty and students, precautions against overreliance, algorithmic bias, and ongoing research ethics.
Some Pakistani universities are beginning to operationalise this policy suited to their academic environment. In its draft framework, LUMS solicits departments to define what it terms AI-fenced competencies, which are the knowledge, judgment and intellectual habits a student must exhibit without using AI, and AI-augmented competencies, those needed in an AI-rich workplace (LUMS Learning Institute,2025). It also recommends delineations between what is considered proper use, misrepresentation, and failure to disclose AI assistance. Agha Khan University adopted a similar approach. It encourages faculty to clearly indicate whether AI is allowed, limited, or prohibited and supports evaluation to focus on the reasoning, analysis, synthesis, and decision-making, rather than the finished product alone (Agha Khan University, 2026).
But, policy statements do not guarantee practice. Recent qualitative research on Pakistani universities describes what Afzal et al., (2026) call moral improvisation. The finding reveals a sheer lack of institutional norms. Because of varying expectations among teachers, courses and peer groups, faculty and students had flexible, real-time decisions to make regarding what was acceptable use of AI.
From Compliance to a Culture of Integrity
This is where the emerging academic integrity networks in Pakistan become relevant. Precisely these questions were the focus of the 2025 Asia-Middle East-Africa Conference on Academic and Research Integrity in Lahore, where universities, policymakers, and specialists in the field of integrity convened. The AMEA-ACARI Consortium, co-founded by Pakistani academic integrity scholar Muhammad Shahid Soroya, has since established its headquarters at the University of Lahore. The accompanying Lahore Declaration advocated for culturally appropriate integrity frameworks and ethical governance and responsible incorporation of new technologies (AMEA-ACARI Consortium, 2025).
This development warrants more attention than another internationally imported model. Academic integrity is in part universal; but the circumstances around which students make academic decisions, are not. The pressures of assessment and faculty capacity, as well as the digital divide and institutional resources and understandings of authorship are different in various educational systems. Thus, Pakistan requires global principles but simply transplanting global templates is not the right approach. (AMEA-ACARI Consortium, 2025).
Some useful policy lessons can be drawn from these precedents. In the age of AI, universities are trying to embrace AI-resistant evidence of learning. That would require assessment reform in tandem with integrity reform. Certain competencies would be secured through viva, supervised exams, practicals, presentations and other skills which require students to think on their feet. Elsewhere, AI could be used more purposefully. For instance, students may be asked to build with AI, critique hallucinated references, compare multiple machine-generated arguments, provide documentation of using an AI model, or indicate bias or discuss why they did not opt for an algorithmic recommendation. This is where the line between assisted and delegated cognition should be drawn.
Pakistan, therefore, must not allow the AI policy to become another compliance document to be signed by students and filed away by universities. Rather, there must be redistribution of responsibility, which is a much harder task. A medical student should have knowledge and judgment that cannot be passed off to ChatGPT. The engineer needs to know how to do the calculations even if the software does the job professionally. A social-science student should be able to build and explain an argument, without outsourcing the reasoning. Those skills involve protected assessment from which authorship can be determined.
Academic integrity has always rested on a bargain. Generative AI has not made that bargain obsolete. It just pushed universities to reassess the perimeters of judging competency. As the next generation of students will be accustomed to using AI as a regular cognitive tool, universities will have to reimagine the academic ecosystem as a whole. Therefore, claims about the success of new AI policies should be treated cautiously. The viable path lies not in prohibition or outright technological surrender. It is to rethink assessment in a way that universities move beyond a futile quest for machine-generated sentences, and to the more significant and deceptively simple question of of whether learning can be verified.
References:
Afzal, A., Rafiq, S., & Oliver, M. (2026). Moral improvisation with generative AI in higher education: faculty and student experiences in Pakistan. Teaching in Higher Education, 1–22. https://doi.org/10.1080/13562517.2026.2695805
Ahmad, Z., Soroya, M. S., & Awais, A. (2026). Understanding academic integrity beliefs, motivations and behaviors amongst postgraduate students: A comparative analysis of public and private sector universities in Punjab, Pakistan. Journal of Academic Ethics, 24, Article 22. https://doi.org/10.1007/s10805-025-09697-x
Aga Khan University. (2026, January 20). Guidance for teaching and learning with generative artificial intelligence. https://www.aku.edu/qtl/programmes/Pages/GenAIuseguidelinesforteachingandlearning.aspx
AMEA–ACARI Consortium. (2025, May 19). ACARI 2025 concludes in Lahore with landmark declaration and handover to Saudi Arabia for 2027. https://www.academicintegrity-uae.com/acariconsortium
Higher Education Commission. (2026). Framework on use of Generative AI (GenAI) tools in higher education institutes (HEIs). https://www.hec.gov.pk/english/policies/Documents/Framework-Gene-AI.pdf
LUMS Learning Institute. (2025). Guiding AI at LUMS: A reflective framework for faculty, staff, and students. https://lli.lums.edu.pk/sites/default/files/2025-06/Guiding%20AI%20at%20LUMS%20-%20Final%20Draft.pdf
Stephenson, R., & Armstrong, C. (2026, March 12). Student generative artificial intelligence survey 2026 (HEPI Report No. 199). Higher Education Policy Institute. https://www.hepi.ac.uk/reports/student-generative-ai-survey-2026/
Shafaq Zernab is a Research Assistant at the Centre for Aerospace & Security Studies (CASS), Islamabad, Pakistan and an MPhil scholar in International Relations at the National Defence University.
The authors' views are their own.
Thank you for being a member of ICAI. Not a member of ICAI yet? Check out the benefits of membership and consider joining us by visiting our membership page. Be part of something great!