Mozambique’s AI-in-Schools Rules: What the New Guidelines Mean

Mozambique’s attempt to govern artificial intelligence in education has progressed beyond the initial announcement made in September. But the latest development is not the adoption of a national AI law.

On 16 September, Education and Culture Minister Samaria Tovela announced that the government was preparing rules for AI use across the education system, with particular attention to generative AI. At the time, the proposal was described as being at an advanced stage of preparation and expected to be submitted to the Council of Ministers before the end of 2026.

The scope initially described was broad: primary, secondary and higher education.

A more recent report, published on 16 October, provides a significant update. The Council Coordinator of the Ministry of Education and Culture reportedly gave a positive assessment to the proposed guidelines, which contain four chapters and 19 articles. They address AI use in teaching and learning, assessment, school management, educational statistics and planning.

That assessment is an important procedural step, but it is not final legal approval. The document still requires refinement, and the latest report says implementation depends on a joint ministerial diploma involving the Ministry of Education and Culture and the Ministry of Labour, Gender and Social Action.

There is also a change in the reported scope. The October account says the proposed guidelines exclude higher education because of institutional autonomy, despite September’s broader announcement. That distinction needs to be resolved in the final instruments and any accompanying official guidance.

As of 22 October 2026, the available sources do not establish that the proposed guidelines have formally entered into force.

The immediate story is therefore about a government trying to turn a broad commitment to responsible AI into practical rules for education, while the details of approval, coverage and implementation remain unsettled.

Education is one of the most visible settings in which generative AI is changing everyday behaviour.

Students can use AI tools to explain difficult concepts, translate material, summarise readings, generate practice questions and obtain feedback on their work. Teachers can use the same technologies to prepare lesson plans, create exercises, adapt explanations and reduce time spent on routine administrative tasks.

Schools can also use AI-enabled systems to organise information and support planning.

These potential benefits explain why a blanket prohibition would be difficult to sustain. AI tools are already widely available through websites, mobile applications and software products. A school policy that ignores them may struggle to reflect how students actually learn and how teachers prepare their work.

But availability is not the same as educational value.

A generated answer may sound authoritative while containing factual errors. A student may submit work that they cannot explain. A teacher may rely on an automated lesson plan that does not fit the curriculum or the students’ needs. A school may adopt a platform that collects more information about children than is necessary for its educational purpose.

There are also longer-term questions about learning itself. If students delegate too much of the thinking process to a system, they may complete assignments more quickly without developing the reasoning, writing and problem-solving skills those assignments were designed to teach.

The opposite is also possible: a student who lacks access to a private tutor could use an AI assistant to receive additional explanations and practise difficult material at their own pace.

Mozambique’s challenge is to establish rules that distinguish these uses by their educational purpose, the level of risk and the degree of human oversight involved.

The government has described the objective as preserving educational quality while ensuring that AI is used responsibly and ethically. The forthcoming guidelines will need to turn that objective into workable expectations for teachers, students, school leaders and technology providers.

Public debate about AI in schools often focuses on academic dishonesty: whether a student used a chatbot to write an essay or whether an assignment reflects the student’s own work.

That is an important issue, but the proposed Mozambican guidelines appear to have a wider scope.

According to the 16 October report, the document addresses AI in five broad areas:

• Teaching and learning: the use of AI to support pedagogy and educational activities.

• Assessment: the role of AI in evaluating students and trainees.

• School management: the use of AI in administrative and operational processes.

• Educational statistics: the handling and production of information used to understand the education system.

• Educational planning: the use of technology to inform decisions about priorities and resources.

The breadth matters because different applications carry different risks.

Using AI to generate optional practice questions is not equivalent to allowing an automated system to determine a student’s final grade. Using a tool to draft an administrative notice is not the same as using predictive analytics to identify students who may be at risk of dropping out.

The latter applications can influence important decisions about individuals. They therefore require stronger safeguards, better-quality data and meaningful human review.

A school that uses AI to analyse attendance or performance data, for example, needs to understand whether the system’s conclusions are reliable and whether students are being unfairly categorised. A student should not be disadvantaged by a system that mistakes limited internet access, language differences or interruptions to schooling for a lack of ability or motivation.

The final guidelines should make clear which uses are permitted, which require additional checks and which should be avoided or prohibited.

The reported four-chapter, 19-article structure indicates an effort to formalise the issue, but the full text of the proposed guidelines was not located in the public sources reviewed for this article. It would therefore be premature to attribute specific obligations, sanctions or detailed technical requirements to the draft.

Generative AI has complicated the distinction between a student’s own work and work produced with assistance.

An essay may be written entirely by a student, drafted by an AI system, translated using a chatbot or developed through a mixture of human writing and AI-generated suggestions. These scenarios are not educationally equivalent.

The same tool can also be used legitimately for brainstorming, explaining unfamiliar concepts, checking grammar or generating practice exercises. Whether a particular use is appropriate depends on the assignment’s purpose and the instructions given by the educator.

A workable policy should therefore distinguish between unauthorised substitution for a student’s work and permitted assistance that supports learning.

That distinction requires clarity from institutions. Students need to know whether AI is permitted for a particular task, what forms of assistance must be disclosed and which parts of an assignment must demonstrate independent understanding.

Teachers need a consistent basis for evaluating the work, rather than having to develop individual rules for every assignment.

The policy also needs to account for the limitations of AI-detection tools. A detector’s conclusion should not be treated as conclusive proof of misconduct without other evidence. Automated systems can make mistakes, and the consequences of an incorrect accusation can be serious for a student’s academic record and trust in the institution.

A fair process should give students an opportunity to explain their work, demonstrate their understanding and challenge allegations. Assessment methods can also be designed to examine learning more directly, for example, by asking students to explain their reasoning, discuss drafts or apply concepts to unfamiliar problems.

For remote learning, these issues are especially important. An institution may not be able to observe a student’s work process directly, but that does not mean it should rely exclusively on surveillance or automated detection.

The strongest rules will make expectations clear before an assignment begins and provide fair procedures when questions arise.

The government’s announcement came during the 4th National Conference on Distance Education, where preliminary data for 2025 indicated that approximately 75,200 students were enrolled in distance higher education, equivalent to 26.6% of the country’s total higher-education enrolment.

The figures help explain why AI governance has become a practical issue rather than a distant policy question.

Distance education can expand access for people who cannot attend conventional classes because of work, family responsibilities, geography or other constraints. Minister Tovela highlighted its role in creating opportunities for young people, women, working adults and others seeking new skills.

Generative AI can support those students by offering explanations, helping them practise a language or providing supplementary study support outside scheduled teaching hours.

Yet remote learning also makes it harder for educators to see how an assignment was completed. Students may be using AI tools without clear institutional guidance, while teachers may have limited time or resources to investigate whether submitted work reflects genuine understanding.

The appropriate response is not necessarily to treat remote learning as a higher-risk activity in every respect. It is to design assessment and disclosure rules that work in both online and in-person settings.

For example, an institution might allow AI-assisted study but require students to identify substantial AI contributions to assessed work. It might use short oral discussions, staged submissions or practical exercises to check understanding where appropriate.

The objective should be to measure learning, not simply to detect the presence of technology.

One important qualification remains: the October report says the current proposed guidelines exclude higher education because of institutional autonomy. The 75,200 figure therefore illustrates the broader context in which AI is being used, but it should not be interpreted as evidence that all those students will be directly governed by these particular guidelines.

The relationship between school-level rules, university policies and any later national AI framework will need to be clarified.

A regulation can establish boundaries, but it cannot make AI useful in a classroom by itself.

Teachers will need to understand what AI systems can do, where they tend to fail and how their use can support the curriculum. They will also need practical guidance on privacy, assessment, bias and the responsible handling of student information.

That is a substantial implementation requirement.

Teachers are not only potential users of AI; they are also responsible for deciding whether an AI-generated explanation is suitable for a lesson, whether a generated exercise reflects the intended learning objectives and whether a student’s work demonstrates understanding.

If the rules focus on restrictions without supporting teachers, implementation may become inconsistent. Some educators may prohibit AI broadly, while others may adopt it extensively without clear safeguards.

Training should therefore be designed around real classroom tasks. Teachers need examples of permitted and inappropriate use, ways to check AI-generated content, and methods for explaining AI expectations to students and parents.

They also need time and institutional support to adapt lesson planning and assessment.

The reported guidelines include the use of AI in teaching, assessment and school management. That breadth makes professional development particularly important: the same educator may be asked to understand AI as a teaching aid, a source of potential academic misconduct and a tool that could affect administrative decisions.

The National Institute of Information and Communication Technologies (INTIC) has separately emphasised the need for a coordinated approach linking legal frameworks, data governance, cybersecurity, digital transformation and capacity development. Its October 1 report also described the responsible-AI initiative at Eduardo Mondlane University.

These developments suggest that AI capability-building is taking place across institutions, although they do not establish the scale of teacher training or the resources that will be allocated to implementing the proposed school guidelines.

The equity problem: will the rules work outside well-connected schools?

A national policy must operate across very different educational settings.

Access to reliable electricity, internet connectivity, suitable devices and digital learning resources can vary between communities and institutions. Teachers may also have different levels of experience with digital tools.

These differences affect what a student can realistically do with AI.

If a school permits AI-assisted assignments but students must pay for a particular service to complete them effectively, the policy could create an unfair advantage for those who can afford it. Students with intermittent internet access may be unable to use the same tools as classmates with reliable connectivity.

A policy that requires extensive digital monitoring could also be difficult for schools that lack the necessary equipment or technical support.

The final rules should therefore distinguish between the educational objective and the specific technology used to achieve it. Students should not be required to purchase a particular AI product simply to meet an assignment’s expectations unless there is a clear, equitable basis for that requirement.

Institutions should also consider offline alternatives, accessible materials and assessment methods that do not assume every learner has continuous access to the same digital services.

Equity matters for teachers as well. Schools with limited resources may not be able to implement complex auditing or monitoring systems. Clear minimum safeguards, practical templates and accessible training can make responsible use more achievable.

This is where regulation connects directly with education policy and public investment. If the government wants AI to improve educational outcomes, it must consider not only which uses are acceptable but also whether institutions have the capacity to implement them.

The public sources reviewed do not yet provide a detailed implementation budget, a national teacher-training plan or an assessment of the infrastructure required for the new guidelines. These are important gaps to watch as the proposal moves towards formal approval.

AI systems used in education may process names, school records, assignments, performance data, attendance information, voice recordings and other personal information.

Some applications may also infer characteristics or generate predictions about a student’s abilities and behaviour. That creates a need to understand what information is collected, why it is necessary, who can access it and how long it is retained.

The risks are particularly significant when children are involved.

A school or teacher may be tempted to paste student assignments, personal circumstances or assessment records into a public AI service to obtain a summary or recommendation. Depending on the service and its terms, this could expose information beyond the school’s intended control.

A responsible framework should therefore address data minimisation, appropriate access controls, security, retention and the use of third-party platforms. It should also make clear when parental involvement or additional safeguards are required.

The legal framework for personal data in Mozambique includes Law No. 3/2017 on electronic transactions and other applicable rules concerning confidentiality, communications and information security. The precise obligations will depend on the type of data, the service and the relevant law. The proposed educational guidelines should not be treated as a substitute for existing legal obligations.

Schools also need to consider whether an AI system makes consequential recommendations about students. If a tool flags a child as struggling, for example, educators should understand the basis and limitations of that assessment before acting on it.

The aim should be to use information to support learning, not to create opaque labels that follow students without a clear explanation or route for correction.

Because the full draft guidelines were not publicly available in the sources reviewed, it remains unclear how specifically the proposal addresses student-data protection, third-party AI providers and retention of information. Those provisions should be examined once an authoritative text is published.

University autonomy is now a central question

The reported change in scope between September and October raises a significant governance issue.

The initial announcement described rules covering all levels of education, from primary through higher education. The October report, however, says the proposed guidelines exclude higher education because of institutional autonomy.

This is not merely a technical drafting detail. Universities have distinct responsibilities for teaching, research, academic standards and institutional governance. A national framework needs to clarify how those responsibilities interact with common safeguards for AI use.

There are arguments for common national principles. Students should be protected from harmful data practices, and expectations around transparency and accountability should not depend entirely on the institution they attend.

At the same time, universities may need flexibility to develop policies suited to different disciplines and research activities. The appropriate use of AI in a programming course may differ from its use in a medical programme, a language examination or a research project.

The reported approach therefore creates a question of regulatory architecture: should the education ministry establish a uniform rule for every institution, or should national requirements set minimum safeguards while universities retain authority over academic implementation?

The answer may involve both. National rules could establish baseline protections and responsibilities, while individual institutions develop more detailed policies suited to their educational missions.

The government will need to clarify the scope of the final instrument, the legal basis for excluding higher education and the relationship between school-level guidance and institutional university policies.

Eduardo Mondlane University’s reported approval of its own AI-use directive provides an example of institution-level policy development. INTIC reported on 1 October that the university had approved a directive intended to guide responsible AI use, with further work planned on responsibilities, resources and monitoring. The report also described the institutional roadmap as still under development.

That experience could inform the wider discussion, but one university’s directive does not establish a common national policy for all higher-education institutions.

The proposed guidelines are being developed while Mozambique is also working on a broader national AI strategy.

INTIC published the first version of the proposed National Artificial Intelligence Strategy for public consultation on 5 May 2026. The invitation sought contributions from public and private institutions, academia, development partners and citizens, with comments due by 4 September.

The strategy is intended to support a responsible, ethical, inclusive and sustainable AI ecosystem, with AI framed as a tool for innovation, productivity and public-service improvement.

But the school-level regulation is advancing on its own timetable. That creates a coordination challenge: educational rules could be finalised before the national strategy is formally adopted, potentially setting precedents that later policy work will need to accommodate.

This is not inherently a problem. Schools face immediate practical questions about AI use, and sector-specific rules can be developed while broader strategy work continues. The risk is that the resulting policies become inconsistent or difficult to reconcile.

Digital-rights organisation CIBERCIDADÃOS raised a related concern in an October 12 policy brief. It argued that public information about the consultation process for the educational rules was insufficient, including information about who had been consulted, which issues had been discussed and how contributions had been incorporated.

The organisation called for wider participation by teachers, students, researchers, educational institutions and civil society, and recommended a transparent roadmap for consultation, review and institutional coordination.

This is an attributed civil-society position, not proof that no consultation took place. The government has said it has held working meetings with educational institutions. The concern is whether the process is transparent enough for the broader education community to understand and influence the rules before the major decisions are finalised.

That distinction matters because teachers and students will be among the people most affected by the final policy. Meaningful participation can reveal practical problems that are not apparent during central policy drafting, including the realities of rural schools, language diversity, unequal connectivity and the costs of compliance.

Consultation is most useful while the policy remains open to change—not after its core choices have already been settled.

Mozambique’s broader AI governance architecture is taking shape

The educational initiative is not being developed in isolation.

Mozambique established a National Commission for Artificial Intelligence (CNIA) through a Council of Ministers decree that came into force in April 2026, according to iAfrica’s reporting. The commission advises the government on ethical, legal, security and social questions involving AI, including its impact on the labour market.

Its responsibilities include providing opinions on policies, strategies, action plans and proposed legal instruments, advising on ethical and security considerations, and supporting international cooperation on AI research and standards. The commission reports to the Council of Ministers.

Meanwhile, INTIC has been developing the National AI Strategy and promoting a broader ecosystem of responsible AI adoption. Its October workshop at Eduardo Mondlane University emphasised the need to connect governance, data, cybersecurity, digital transformation and capacity development.

The emerging picture is of several connected layers:

• A national strategy intended to guide the country’s longer-term AI development.

• A national commission providing policy and technical advice.

• Proposed education guidelines focused on practical uses of AI in teaching, assessment and school management.

• Institution-level policies, including the initiative reported at Eduardo Mondlane University.

The challenge is ensuring that these layers complement one another.

If the education guidelines establish requirements that differ from the national strategy or other future instruments, schools and universities may face conflicting expectations. If the responsibilities of the education ministry, the AI commission and relevant technology authorities are unclear, implementation may be inconsistent.

A coherent framework would define which institution sets policy, which provides technical guidance, who monitors compliance and how the education sector’s experience feeds back into national AI governance.

The proposal’s practical value will depend on the detail of the final instrument. Several issues deserve particular attention.

Clear definitions of permitted and prohibited use. Students and teachers need to know when AI can be used for support, when its use must be disclosed and when it undermines the purpose of an assessment.

Proportionate safeguards. A low-risk study aid should not automatically face the same restrictions as an AI system making consequential decisions about a student. Requirements should reflect the potential harm.

Human responsibility. Schools should know who is accountable for checking AI-generated material and reviewing decisions influenced by automated systems. AI should not become a way to avoid responsibility for educational decisions.

Privacy and child protection. The rules should address personal information entered into AI tools, the use of external providers, retention of data and safeguards for children.

Fair assessment procedures. Institutions need a fair way to investigate suspected misconduct, including an opportunity for students to explain their work. AI-detection results should not automatically determine guilt.

Teacher training and support. Implementation should include accessible guidance, examples and training that can be used by teachers with different levels of digital experience.

Equitable access. The rules should not assume that all students have reliable internet access, personal devices or the ability to pay for premium AI tools.

Coordination with higher education. The final instrument should explain its scope and how it interacts with university autonomy and institution-level AI policies.

Transparent implementation. The government should clarify the final approval process, publication of the authoritative text, responsibilities for oversight and any review mechanism for the policy.

These are analytical priorities rather than claims that the current draft already contains each safeguard. The publicly available reporting does not yet provide the full text needed to assess the provisions article by article.

What schools, universities and technology providers should do now

Until the final guidelines are formally approved and published, institutions should distinguish between existing obligations and preparation for proposed requirements.

Schools and universities can begin by mapping the AI tools already used by teachers, students and administrative staff. This should include both officially approved platforms and tools that individuals may be using independently.

They can then establish interim guidance covering:

• Which AI uses are allowed for each type of assignment.

• When students should disclose substantial AI assistance.

• How teachers should review AI-generated educational material.

• What personal information must not be entered into external tools without appropriate authorisation.

• How suspected academic misconduct will be investigated fairly.

• Who is responsible for approving and monitoring AI applications.

Technology providers selling to educational institutions should be prepared to explain how their products work, what information they collect, how data is protected and what limitations apply to their outputs.

Institutions considering AI for assessment, student support or school management should also evaluate whether the system is suitable for the relevant educational context. A product marketed as an AI tutor or analytics platform should not be assumed to be effective simply because it uses advanced technology.

The final policy may introduce additional obligations. The immediate goal should be to establish good practice while avoiding unsupported assumptions about rules that have not yet entered into force.

Mozambique’s proposed classroom AI guidelines are becoming more concrete. The document has reportedly received a positive assessment from the Ministry of Education and Culture’s coordinating council and is moving through further refinement towards formal approval.

The proposal addresses a genuine policy challenge. Generative AI can support teaching, help students learn independently and improve administrative work. It can also weaken assessment, expose student information, reproduce errors and create unequal advantages when access to technology differs.

The government needs to turn broad commitments to responsible and ethical use into rules that are clear enough to apply consistently and flexible enough to reflect different educational settings.

The latest developments also reveal unresolved questions: the final legal status of the guidelines, their relationship with higher-education autonomy, the transparency of consultation and the resources available for implementation.

Mozambique’s wider AI strategy and national commission provide a context for this work, but the school-level instrument will need to coordinate with those initiatives rather than operate as a disconnected set of requirements.

The success of the policy will not be measured by how many restrictions it introduces. It will be measured by whether teachers can use AI responsibly, students can learn without being unfairly disadvantaged, and educational institutions can protect rights while improving the quality of teaching and learning.