This report captures the main discussion, emerging insights, and practical next steps from the AI in Education Roundtable on Safeguarding and AI Literacy held on 8 April 2026. The session brought together education practitioners from multiple LMIC contexts to reflect on how AI is entering their communities, what opportunities it creates for teaching and learning, what safeguarding concerns it raises, and what kinds of support organisations need to move forward.
Session context
The roundtable was designed around two closely connected themes: AI literacy and child safeguarding in AI contexts. The presentation used in the session framed both as core issues for schools and education organisations working with children in a period of rapid AI adoption, particularly in contexts where infrastructure, policy, teacher preparation, and child protection systems are still evolving.
Across the plenary, breakout contributions, and notes, one message was consistent: AI literacy and safeguarding cannot be treated as separate agendas. Participants repeatedly suggested that schools need to think about capability and protection together, since children and educators are already encountering AI tools in classrooms, assessment, administration, and everyday digital platforms.
AI literacy: opportunities, challenges and solutions
Participants described AI literacy as both a necessary area of investment and a source of widening inequality if schools are unable to respond deliberately. While there was enthusiasm about AI's potential to support teaching and learning, there was equally strong concern that most systems are still underprepared.
Opportunities and challenges by context
Organisations shared opportunities and challenges they are currently facing in their contexts. Some of these include:
Mexico: AI literacy is not yet part of the general curriculum, which contributes to resistance from schools to introduce it formally. Participants noted that students are already experimenting with AI, but many teachers are not adopting it or are unsure how to evaluate AI-generated outputs.
Kenya, Mozambique, Bangladesh, India: Several teachers are struggling to shift toward AI-aware assessment. Participants worried that students are losing confidence in their own critical and creative thinking as AI tools become more available, and raised concern about a new AI-driven divide between schools that can adapt and those that cannot.
Somaliland: Limited access to technology and uneven understanding of how to use AI effectively remain major barriers. Participants emphasised that AI should help learners and teachers build on their knowledge and skills, not replace creativity or independent thought.
Zambia: Educators often lack AI knowledge, skills, and confidence, while poor internet service limits meaningful experimentation. There was also concern that some teachers may overuse AI in planning without enough pedagogical judgment.
Pakistan: AI literacy is not yet embedded in existing ICT courses. Organisations are planning curriculum upgrades and teacher training, but unstable internet, limited devices, and uneven familiarity with AI make implementation difficult. Participants also observed that students are often more willing to experiment with AI than teachers.
Afghanistan and Pakistan: Low digital access and literacy make it hard to build AI meaningfully into assignments. Shared devices create privacy concerns, and uneven literacy means some students use AI without understanding while others do not use it at all.
India: AI is not yet fully embedded in school curricula, and there is no clear cadre of dedicated AI-Computer Science teachers. At the same time, participants identified major opportunities in developing standardised curriculum frameworks and aligning competencies, learning outcomes, and assessments more coherently.
Across multiple LMIC contexts: Participants repeatedly highlighted device scarcity, limited connectivity, weak foundational literacy, gaps in pre-service teacher training, and low understanding of how AI systems work. At the same time, they saw promise in unplugged approaches and integrating AI literacy across subjects rather than confining it to standalone ICT lessons.
Professional development on AI use and ethics for teachers, alongside learner-facing sessions on safe and effective use of AI.
“AI champions” networks to support peer learning and help teachers build confidence, especially around ethics and bias.
Early pilots with AI-enabled learning tools, including Khanmigo for teachers in Pakistan and AI tutoring or literacy support models in different settings.
Curriculum design work that maps AI competencies to learning outcomes and assessment methods, including efforts to move from ad hoc experimentation to a more structured progression.
Assessment redesign so that schools pay attention not only to final products but also to how students use AI in the learning process, including prompting, reflection, and judgment.
Safeguarding and AI: concerns, opportunities and solutions
AI has introduced new safeguarding risks that existing school policies were never designed to address. In LMICs, limited digital governance, weaker platform accountability, and low AI literacy among educators compound each threat. Organisations shared concerns, opportunities and solutions they are currently exploring in their AI journey:
Safeguarding concerns by context
Cross-context data privacy concerns – Participants raised concern about direct student contact with AI tools that collect voice, photos, IDs, and other sensitive information, often through platforms that already sit inside everyday school or family communication channels. There was concern that many users do not fully realise what data is being shared.
Afghanistan and fragile settings – In conflict-affected or politically sensitive environments, participants stressed that personal data exposure can have serious offline consequences for students and families, making privacy and secure practice especially urgent.
Across multiple contexts – Participants identified risks such as unsafe or contextually inappropriate AI responses, emotional manipulation, overreliance on AI, academic integrity concerns, and unhealthy dependence on AI tools. They also questioned the long-term effects of excessive AI use on cognition, creativity, and children's relationships with knowledge.
School leadership and teacher practice – Low teacher confidence and limited training make it difficult for adults to identify and respond to AI-related risks, including misleading outputs, unsafe recommendations, and misuse of student data.
Equity and exposure – Participants worried that unequal access to safe, well-governed AI tools could deepen educational inequality, with some children benefiting from high-quality tools and others either excluded or left with unsafe often free alternatives.
Solutions participants are exploring included:
AI guardrails and school or network-level policies that clarify approved tools, acceptable uses, and review processes before adoption.
Simulations and scenario-based learning activities that help staff and students think through AI-related safeguarding risks in concrete ways.
Child-safe AI creation tools and more bounded use-cases that reduce open-ended exposure to high-risk tools.
Tool-vetting processes, including the use of external frameworks before adoption, so that privacy, contextual relevance, and safeguarding are considered alongside pedagogical value.
Digital behaviour modelling and mental health safeguarding lessons, so that responsible AI use is framed as part of a broader culture of care and trust.
Support participants said they need
The post-session survey provided a practical picture of what organisations want next. Participants are looking for help that is immediately usable in school and organisational settings.
Survey respondents asked for:
Tools, resources, and training on how to use AI meaningfully with educators and students.
AI literacy training for educators, plus examples of effective and responsible school use.
Guidance on using AI to support literacy and numeracy, and practical advice on how teachers can integrate these tools into curriculum.
AI safety guidelines for students and access to student-safe AI tools.
Researched AI pedagogy, AI assessment frameworks, and unplugged AI activities that work in lower-resource settings.
Support on where to start with AI while maintaining integrity, equity, creativity, privacy, and trust.
- Sensitisation around AI use, data sharing, and privacy risks.
The survey also suggests that organisations are already taking some initial steps. One respondent described an “AI Playground” for educators and a working group researching how to embed AI literacy into curriculum, while also reviewing AI products and interventions globally. That indicates a strong appetite for peer learning and shared evidence.
Key insights
In summary, these key insights were reflected across the conversations:
Teacher readiness is central: Whether the issue is AI literacy or safeguarding, participants kept returning to teacher confidence, judgment, and practical support as the hinge factor in responsible implementation.
Equity shapes every conversation about AI in education: Access to devices, connectivity, safe tools, and contextually relevant training remains highly uneven, and participants fear AI could widen existing divides if these inequalities are ignored.
Safeguarding gaps are critical: Participants framed privacy, emotional wellbeing, academic integrity, and child-safe design as issues that must be built into AI adoption from the beginning.
There is strong demand for practical guidance: Organisations are seeking models, examples, policies, tools, and decision frameworks they can adapt to their own settings.
Next steps
We are continuing the discovery sessions of our AI in Education Roundtable Discussions. We are inviting members and partners to help shape the next phase of this work. You can contribute in several ways:
Join upcoming roundtable discussions
Take part in thematic working groups and learning sessions
Share your organisation’s experiences, questions, or tools
Stay informed and contribute feedback as this work evolves
AI and Instructional Design (12 PM UK, April 29 2026)
- AI and Assessments (12PM UK, May 13, 2026)
Your input will help shape how the GSF community navigates the responsible use of AI in education, strengthening teaching, protecting trust, and supporting better learning and futures for children.
