.
The writer is a Visiting Research Fellow at Middlesex University, UK. He can be reached at naveed.r.khan@gmail.com
The recent initiatives from the Sindh and Punjab provinces to introduce AI in government schools for teachers and students mark an important moment for Pakistan’s education system. This is the first time AI is entering schools at scale. Yet this bold step also reveals a critical gap. Pakistan is deploying AI in education without a dedicated governance policy that outlines how such technology should be assessed and held accountable.
Pakistan’s National AI Policy 2025 provides an extensive, economy-wide vision for innovation, infrastructure and talent development. But education is not just another sector. It involves children, public trust, learning outcomes and long-term social consequences. A general AI policy, by design, cannot appropriately address these complexities. Thus, initiatives by Sindh and Punjab need an urgent standalone AI-in-Education Policy.
Global experience strengthens this argument. For instance, Unesco’s global framework on AI in education stresses that AI literacy is not just about using tools, but about understanding ethics, bias, transparency and human oversight. Unesco advises that education systems adopting AI without a clear governance framework risk increasing discrepancy, weakening teacher autonomy and compromising student rights. Notably, only a small number of countries globally have integrated AI learning objectives into national curricula; those that have done so treat governance as foundational rather than optional.
The US offers a cautionary parallel. The US Department of Education has emphasised that AI systems progressively shape decisions about assessment, feedback and student profiling. Without clear rules, these systems can introduce bias, privacy violations and ambiguous decision-making into schools. The US response has focused on accountability, human-in-the-loop safeguards and the right to challenge automated decisions. Their message is clear: once AI impacts educational outcomes, it becomes a governance issue, not merely a technological one.
The UK has gone further in operationalising this insight. Its policy direction for AI in education prioritises safety, transparency and teacher control. Schools are advised to treat AI outputs as inherently unreliable unless verified, to avoid automated grading without human judgment, and to ensure procurement standards include auditability and data protection. The UK permitted the adoption of AI within clearly defined boundaries.
Pakistan presently lacks such boundaries.
The most under-discussed gap is procurement. As provinces and institutions acquire AI tools, often from private vendors, these systems begin to shape how students learn, how teachers teach, and how performance is measured. Yet there are no national standards governing bias testing, data retention, model updates or mechanisms for appeal when AI-assisted decisions cause harm.
This gap matters now because education systems scale by default. What begins as a provincial initiative can quickly become rooted nationwide. If early deployments regulate weak safeguards, reversing course later becomes constitutionally and institutionally difficult. Moreover, public criticism of surveillance, unfair assessment or academic integrity could weaken trust in AI and in education reform.
Pakistan is struggling with graduate employability and skills mismatch. AI in education, if poorly governed, may worsen this mismatch by prioritising tools over thinking, automation over pedagogy.
A Pakistan AI-in-Education Policy would not slow innovation. It would enable it responsibly. At a minimum, such a policy should establish three principles. First, teachers before tools: mandatory training and clear human oversight rules must precede student exposure. Second, learning outcomes over deployment metrics: success should be measured by capability development, not by the number of AI platforms rolled out. Third, accountability by design: procurement standards must require transparency, audit rights, data protection and redress mechanisms.
Sindh and Punjab initiatives can become a national strength, albeit only if they are treated as test cases rather than templates. We need to do what leading systems have learned the hard way — build governance first, capability second, and technology last.
