WSIS Action Line C7 E-learning

Mark West argues that public digital learning platforms should function as digital commons for education. In his framing, these platforms give learners, teachers and families access to quality-assured, curriculum-aligned and age-appropriate resources so that education remains a public good and a human right in digital spaces as well as physical ones.

He explains that the Gateways Initiative helps countries build digital commons for education where learners, teachers and families can access quality-assured, curriculum-aligned and age-appropriate learning materials . He also explicitly states that this is tied to the idea that education is a public good and a human right in digital as well as physical and social realms .

Mark West argues that openness should emerge from the technical and institutional design of digital learning systems, not from superficial compliance measures added later. His point is that if the underlying infrastructure is built well, openness follows naturally and can be shared as a practical model across countries.

In reflecting on Benya's presentation, he highlights her conclusion that 'if we get the infrastructure right openness is inevitable' and says this shows the value of the Gateways Initiative in helping countries understand what that infrastructure should look like . He also says the artist example demonstrates a practical way to ensure openness despite copyright limitations, turning an abstract concept into something operational .

Mark West clarifies that the UN agencies are not creating a single global education platform. Instead, they support countries in building their own national or subnational systems, some of which already include AI functions depending on each country's context and progress.

When asked about AI enablement, he states directly that the United Nations does not run a platform and instead helps countries build their own national or subnational platforms . He adds that they are not building a global platform because education platforms should be controlled at national and subnational levels, while noting that many countries already have AI-enabled platforms and others are just beginning that process .

Mark West argues that legal pressure from publishers should not undermine the availability of public and open educational content. He maintains that free and accessible learning materials are essential to fair educational use and should remain part of digital commons for education.

In response to the copyright question, he notes that Oak National Academy is also being sued by publishers and remarks that lawsuits are occurring in multiple places . He nevertheless insists that there must be a place for free, open and public learning content accessible to all, and describes that as the foundation of fair use .

Mark West argues that cybersecurity is already covered within the Charter's broader trustworthiness framework. He presents safety and security as integral to the guidance being offered, rather than as a separate or optional concern.

When asked whether cybersecurity recommendations are included, he responds affirmatively and says that the trustworthiness principle is about the safety and security of the platforms . He also indicates that this area is connected to the technical work that Carolina could elaborate on .

Frank Van Capelle presents the Charter as a landmark global framework for public digital learning platforms. He argues that it is the first time the international community has collectively set out norms for how such platforms should be governed, designed and developed.

He states that the Charter is 'the first time ever' that the global community has come together to create a normative framework around public digital learning platforms . He adds that this framework was built through participation by member countries, three UN agencies, experts and wider public consultation .

Frank Van Capelle argues that openness must be designed into digital learning platforms from the start. In practical terms, this means avoiding vendor lock-in, using open standards, enabling interoperability and allowing data exchange across tools and systems.

He explains that openness means building platforms so they do not create vendor lock-in, and so they use open standards, remain interoperable and can integrate with other solutions . He also broadens this by pointing to the wider EdTech ecosystem, stressing that platforms should work side by side with other tools rather than remain siloed and unable to exchange data or be easily replaced .

Frank Van Capelle argues that educational platforms and AI tools must be designed with pedagogy at their core. He warns that generic tools, including general-purpose chatbots, can be harmful when applied to education without being purpose-built for learners and teachers.

He says the pedagogical principle requires platforms to be anchored in education and notes that many tools used in education were not built for that purpose . As an example, he cites generic chatbots and says they can cause great harm because they are not designed for education .

Frank Van Capelle argues that multilingual inclusion is a core part of inclusive public digital learning platforms. While AI creates new opportunities to translate materials into both official and minority languages, he stresses that the outputs still require human quality assurance because fully automated translation remains unreliable.

He explicitly links languages to the inclusion principle and says digital platforms can provide content in official languages and increasingly in minority languages . He adds that AI now makes it possible to begin translating even into small languages, but says live on-the-fly translation does not yet work well enough, so a quality-assurance layer is still needed . He gives the example of an AI tool called Accessible Digital Textbooks that converts materials into different languages and formats accessible to children with disabilities .

Frank Van Capelle argues that trustworthiness is a core design requirement for public digital learning platforms. He defines this in terms of reliability, safety and transparency, and links it to the public role in overseeing and safeguarding these systems.

In outlining the Charter's principles, he says platforms must be 'trustworthy, reliable, safe, transparent' and that this is an area where the public role is especially important . He also frames public management and ownership as helping ensure systems are designed for all children, including those otherwise excluded .

Zeynep Varoglu argues that openness in digital public learning platforms has several concrete dimensions rather than being a vague principle. She identifies interoperable systems, modular architecture, open licensing and open-by-design access as the core elements needed to make platforms genuinely open and accessible.

She introduces four points from the Charter's section on open: connected interoperable systems, modular architecture, open licensing and open by design . She explains that this means systems should work together intuitively, avoid vendor dependence for reasons of information and knowledge sovereignty, openly license content, and reduce unnecessary barriers such as forced log-ins for learners who simply want access .

Zeynep Varoglu argues that openness must be understood as an ecosystem, not just a single policy on content. In her view, digital learning platforms should combine openly licensed content, software and data, and also connect to broader open-access practices in knowledge production.

She says UNESCO's work is about creating an ecosystem of open so that information becomes accessible and grows exponentially . She specifies that this includes openly licensed content, software and data, and also mentions open access and especially diamond open access to research, where there are no fees for authors or users .

Zeynep Varoglu argues that OER Commons is a useful and respected platform within the broader open education landscape. She presents it as a strong partner and suggests that systems should make use of OER resources as part of larger open digital learning architectures.

Responding to the audience question, she says that all systems are invited to use OER and that OER Commons is 'a very good system' and 'a very good platform' as well as a close partner of UNESCO . She also notes that it would be useful to examine what can be used from OER Commons on its side within this wider architecture .

Zeynep Varoglu argues that AI-based translation is not only a technical challenge but also a legal one. She points out that licensing issues remain intertwined with the use of AI for translation, so multilingual access cannot be treated as purely a matter of model performance.

In answering the question about translation and LLMs, she says there is also an issue of licensing that arises and describes it as 'a tricky one' . She adds that this concern comes up particularly when AI is used for translation .

Zeynep Varoglu argues that open educational resources have strong legitimacy because they are backed by international normative agreements. Her point is that even when publishers resist, OER remain supported by formal commitments from UNESCO member states and governments.

She reminds the audience that 194 countries adopted normative instruments supporting OER in educational systems . She further emphasises that this is a UNESCO and UN-backed instrument, and says OER therefore have national and governmental support at the UN level .

Carolina Anselmino argues for differentiated governance across layers of education technology systems. She says sector-level systems require stronger government control because of sovereignty, compliance and security concerns, while school-level systems may be more distributed but must remain interoperable and aligned with government frameworks.

She explains that systems at the education sector layer are strategic, require high compliance and security, and therefore should be under greater government control because they involve more sensitive data and issues of sovereignty . She contrasts this with school-level systems, which still need to interoperate and share data back with other layers while complying with government frameworks even if they are not directly government-controlled .

Carolina Anselmino argues that the Charter now needs to move from normative guidance to implementation tools. She describes a second phase focused on producing technical specifications, reference implementations and capacity support so countries can procure or build platforms that embody the Charter's principles from the outset.

She says the work is entering 'a second phase' and describes it as moving 'from principle to practice' or operationalising the Charter . She explains that this involves translating principles into technical specifications and reference implementation materials to support countries procuring or building digital learning platforms with those principles embedded by design . She also notes that three working groups have already been created on content repositories, registries and analytics/verifiable credentials .

Carolina Anselmino argues that educational content should not be viewed merely as static material such as digitised textbooks. Instead, it should be built and structured as infrastructure, accessible through APIs so others can create translations, new formats and additional innovative services on top of it.

In response to the OER and translation discussion, she says there was a shift in framing towards looking at content as infrastructure . She explains that when content becomes infrastructure through APIs, it can be used to build translations and other innovative materials and services on top of it . She also contrasts this with simply digitising textbooks, arguing instead for building content and curriculum as infrastructure .

Carolina Anselmino argues that AI should not be conceptualised as a separate module within digital learning systems. Instead, it is a capability that cuts across and becomes embedded in multiple parts of the platform ecosystem.

She states directly that AI is 'not a building block by itself' but is instead 'a cross-sectoral capability' embedded in the digital learning platforms now being developed and observed .

Benya argues that openness can be operationalised through deliberate curriculum and database design choices. At Oak Academy, this meant creating lessons that are third-party-content-free wherever possible and tracking restricted material at component level so only the restricted parts are limited, while the rest remains open.

She explains that many Oak lessons are 'third-party content free by construction', which allowed broader cultural diversity than would have been possible with licensed content . She also says Oak moved from tracking third-party content at whole-lesson level to component level, so that only specific restricted elements are blocked while items such as slide decks, worksheets and quizzes can remain open and available through APIs . She adds that achieving this required reworking database schemas, rebuilding authoring tools and changing teacher creator workflows .

Benya argues that precise ontology and data modelling are what make controlled openness practical. By encoding curriculum structure and restrictions at a granular level, Oak can expose much of its content openly through APIs while still keeping the full teaching experience usable in classrooms.

She says that when the ontology and data model are correct, Oak can control what it does and does not surface . She explains that teachers can access 100% of the content in classrooms while about 80% can be surfaced through APIs under an open government licence, because the ontology encodes curriculum logic and restrictions precisely . She adds that the same structure supports adaptable lesson responses and allows AI systems to draw on misconceptions and prior learning even where full content cannot be directly surfaced .

Benya argues that AI works better in educational settings when curriculum intelligence is built into the underlying ontology. If the system captures lesson logic, misconceptions and prior learning, AI can generate more reliable and context-sensitive adaptations even when it does not have access to every piece of content.

She explains that because Oak's curriculum thinking is baked into the ontology, an AI tool plugged into the Oak MCP can still understand what makes a good lesson, what misconceptions matter and what prior learning is required, even if the API cannot access the entire curriculum . She gives the example of lessons designed in a template for local adaptation, such as switching a London example to Manchester, and says the same structured logic can support AI-driven adaptation .

Benya argues that copyright remains a real constraint on open educational platforms, but its effects can be mitigated through careful technical and pedagogical design. By minimising reliance on third-party material and isolating restrictions to the smallest possible components, platforms can preserve wider openness and usability.

She says Oak began from a desire to uphold openness while 'wrestling with licensing limitations' and developed component-level data tracking and repeatable design rules in response . She also notes that restrictions and limitations led Oak to better pedagogy and better lesson design because they had to move away from reactive dependence on licensed materials . Earlier, she explained that component-level restriction means only a specific slide or element may be blocked, while most of the lesson remains open and accessible through APIs .

The audience member argues that interoperability with OER platforms is an important practical issue for open learning ecosystems. The question suggests that mixing and matching resources across languages and metadata systems should be possible, but that some existing OER platforms may not yet have the infrastructure needed to support this well.

Kaveh Bazargan asks whether it is possible to mix and match with Open Educational Resources and refers specifically to OER Commons as a platform with different languages, metadata and ontology . He also suggests that OER Commons may not have the same kind of infrastructure being discussed, and asks how multilingual support would work in that context .

Kamil Zubek argues that current large language models are still weak at translating highly specialised scientific language. He warns that this creates delays in making knowledge available in other languages and therefore widens educational gaps between countries.

He states that large language models are 'not very good' at translating specific and detailed scientific language and that this creates errors or gaps in translation . He also explains that although textbooks in English may exist, translating them into other languages takes time and creates knowledge gaps for countries depending on how long translation takes .

The audience member raises a practical implementation question about what AI-enabled educational platforms can actually do. In particular, the question focuses on whether such platforms support personalised functions such as learning pathway recommendation.

The speaker asks how AI-enabled the platform is and references prior involvement in a United Nations project in Vietnam on learning pathway recommendation . He then asks directly whether the platform being discussed can perform that kind of recommendation function .

Stephen Wyber argues that the education community may need to advocate more strongly for copyright rules that support educational access and openness. He frames this concern against a backdrop of litigation by publishers and broader pressure for tighter copyright control, including in AI-related debates.

He refers to copyright as a live issue and notes that AI debates elsewhere may be pushing for stricter rules . He gives examples of French educational publishers taking a region to court for putting open educational resources online and mentions the upcoming WIPO General Assembly as a relevant policy moment . He then asks whether the sector needs to be 'noisier' to ensure copyright frameworks actually work for education .

Christine Housel raises a practical implementation question about whether cybersecurity is built into the support and recommendations offered for these platforms. The question highlights security as a necessary operational concern alongside openness and accessibility.

She asks directly whether recommendations around cybersecurity are part of what the initiative offers in its development process .