The Ministry of Education, Science and Technology recently launched the National Guidelines for Artificial Intelligence in Education, a framework that forms part of the broader National Digital Education Strategy 2024/25–2029/30.
The document arrives at a critical juncture, as artificial intelligence steadily reshapes service delivery across sectors and slowly makes its way into Tanzanian classrooms.
Whilst the guidelines aim to strike a balance between promoting responsible AI use and safeguarding traditional learning values, they risk becoming another unimplemented framework if structural dependencies are not addressed.
The current debate centres on how to harness AI’s potential without compromising the development of students’ critical thinking skills. Education experts warn that overreliance on AI can dull intellectual curiosity, as struggling with a concept helps develop essential reasoning skills.
When that struggle is skipped through AI assistance, a critical part of learning is lost, risking the creation of a generation that can follow instructions but not innovate.
To address this, the government is prioritising AI literacy in school curricula, ensuring both teachers and students understand the ethical, social, and cognitive implications of its use.
The guidelines recommend a blended learning approach where AI supports rather than replaces human teaching and reasoning. However, the true test lies not in the policy documents, but in the architecture of the systems we choose to deploy.
Open source foundations
The guidelines need an unambiguous requirement: any AI system approved for educational use must be built on open-source foundations. Data must be stored on Tanzanian soil, and models must be fine-tuned on locally curated datasets owned by Tanzanian institutions.
International collaborations should be welcomed only if they transfer capacity and code, rather than just product licences.
We must grow a local AI developer ecosystem instead of perpetually renting black-box APIs from abroad. If the government has to put skin in the game to avoid complete dependency, then so be it.
The alternative is a form of digital colonialism by choice, where proprietary systems owned by foreign vendors store our children’s data on servers we will never touch.
READ MORE: Tanzania Court Adopts Artificial Intelligence (AI) in its Processes
The proposed digital monitoring, evaluation, and learning system is conceptually sound, but without enforcement mechanisms, it risks becoming another bureaucratic exercise. Fortunately, we already have institutions with legal authority to oversee this implementation.
Ethical monitoring structures
This monitoring should be added to the purview of educational inspectors at the district level and the Controller and Auditor General’s office at the national level. Their findings should feed into annual curriculum reviews by mandate, serving as a standing agenda item with an obligation to respond.
Before auditing begins, the criteria must be co-developed by the national audit office alongside educators, students, and community representatives.
This collaborative approach prevents the exercise from becoming a top-down checklist that misses classroom realities entirely. Whilst fraud and misallocation may persist, and audit reports may be ignored by those who find them inconvenient, this architecture creates an unerasable record and a clear trail of accountability failure. In a high-entropy system, that trail is the first prerequisite for any corrective action.
Additionally, the guidelines should mandate that every school deploying AI establish a governance committee of teachers, students, and parents. These committees must possess real power to reject a tool or demand local adaptations, ensuring that the ethical architecture is not top-down and brittle.
No reverse gear
A former education minister once had to intervene to reverse disastrous curriculum experiments, an action possible because a curriculum is ultimately a document that can be reprinted.
READ MORE: From Scaling Up to Integrating Tech-Centric Interventions in Tanzania’s Schools
AI in education, however, has no reverse gear. If we deploy frictionless answer-machines before children build the metacognitive muscles to resist intellectual atrophy, the consequences will not be fixable by a future minister’s circular.
The cognitive habits will be formed, the data extracted, the local developer ecosystems stillborn, and the trust between teachers and the ministry severed entirely. These are compound-interest losses that accrue silently until we have a generation that cannot think without a prompt.
History will remember how Tanzania handled this moment, not whether we merely issued a guideline. History will remember whether we had the rigour to say no to bad AI, the patience to build small and offline-first, and the humility to ask teachers rather than inform them.
The Ministry’s guidelines correctly identify that AI can either widen the chasm between the haves and the have-nots or become a bridge. We are deciding right now whether a generation of Tanzanian children will use AI as a cognitive crutch that atrophies their reason, or as a scaffold that strengthens it.
Godbless Baluhya says he likes beer and technology. He can be reached at godbless.baluhya@protonmail.com. The opinions expressed here are the writer’s own and do not necessarily reflect those of The Chanzo. If you are interested in publishing in this space, please contact our editors at editor@thechanzo.com.
