The University of Manchester

A dialogue is essential for driving the discussion forward. However, what follows after the dialogue is equally important: compiling questions and themes to guide researchers and practitioners, identifying best practices to share, exploring the root causes of poor practices, and establishing effective communication channels for a sustained network. This will help keep the conversation going and support the development of practical solutions.

• Safe, secure and trustworthy AI
• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight

Currently, there are many open-source software options and both free and paid AI tools available. However, governance and support for AI capacity building are lagging behind. This has resulted in limited understanding not only of the technical aspects of AI, but also of its social dimensions and its impact on society. Without adequate support, there is a risk of widening the existing digital divide, creating new forms of digital exclusion, or even contributing to what might be described as "AI poverty." Therefore, I have selected the four aspects above as priorities.

One important issue not fully captured by the listed themes is AI inclusion and accessibility. Generative AI has significant potential to improve accessibility, although these tools are not officially classified as assistive technologies. Research has explored the use of AI with marginalised learners, including learners with disabilities (Zhao et al., 2025) and learners in Global South contexts (Zhao et al., 2026). These projects highlight how access to AI is shaped not only by individual skills, but also by wider structural barriers such as digital infrastructure, platform availability, language representation, and the design of AI tools and guidance. In particular, many AI systems are developed around English-dominant datasets, interfaces, instructions and assumptions about users. This can exclude learners whose languages, cultures, educational contexts, or accessibility needs are not adequately represented in training data or tool design. For learners in low-resource settings, AI tools may also be inaccessible because of poor connectivity, cost, device limitations, or lack of locally relevant support. Therefore, AI inclusion and accessibility should be treated as a cross-cutting issue. It connects technical design, linguistic diversity, disability rights, digital inequality, education, and social justice. Without this focus, AI risks reproducing existing exclusions rather than supporting more equitable participation. References: Zhao, X., Miao, F., Xie, H., & Chen, X. (2026). Exploring Student and Educator Challenges in AI Competency Development: A Comparative Analysis. Multimodal Technologies and Interaction, 10(3), 27. https://doi.org/10.3390/mti10030027 Zhao, X., Cox, A., & Chen, X. (2025). The use of generative AI by students with disabilities in higher education. The Internet and Higher Education, 66, 101014. https://doi.org/10.1016/j.iheduc.2025.10101

The governance gaps and related developments in AI are affecting the UK, China and Global South contexts in different but interconnected ways, particularly in the higher education sector. In the UK, universities are developing policies and guidance, but approaches remain uneven across disciplinary and institutions. There are opportunities to use AI to enhance feedback, accessibility, administrative efficiency and inclusive learning design. However, there are also concerns about commercial AI platforms, data privacy, bias, hallucination, and how to ensure staff and students, especially those marginalised learners (Zhao et al., 2025a; 2026a), have the capacity to use AI critically and responsibly. In China, AI development is closely linked to national innovation, digital education and technological competitiveness. There are strong opportunities for large-scale AI adoption in education and top down policy guidelines. At the same time, governance challenges include balancing innovation with transparency, ethical use, human oversight, and the protection of learners' agency, especially some common GenAI tools are restricted due to geopolitical reasons (Zhao, et al., 2025b). In Global South contexts, such as Africa, and southeast Asia, the most significant issue is uneven access. AI tools are often designed around English-dominant datasets, infrastructures and assumptions, which can marginalise users whose languages, cultures and educational systems are underrepresented. Limited digital infrastructure, cost, and platform access may further widen existing inequalities, risking training remaining theoretical rather than practical (Zhao, et al., 2026b). Across these contexts, AI capacity-building is essential. Teachers and students need more than technical skills; they need critical, ethical and pedagogical competencies to evaluate AI outputs, understand limitations, and make informed decisions. The key opportunity is to develop more inclusive, multilingual and context-sensitive governance approaches that support trustworthy AI while recognising different cultural, economic and educational realities. This would help ensure that AI benefits are shared more equitably rather than reinforcing existing global inequalities. References: Zhao, X., Chen, X., & Cox, A. (2026a). Exploring the affordances of generative AI in academic writing for students with disabilities: A bottom-up approach to inform GenAI policies. Policy Futures in Education, 24(1), 59-81. https://doi.org/10.1177/1478210325139 Zhao, X., Miao, F., Xie, H., & Chen, X. (2026b). Exploring Student and Educator Challenges in AI Competency Development: A Comparative Analysis. Multimodal Technologies and Interaction, 10(3), 27. https://doi.org/10.3390/mti10030027 Zhao, X., Cox, A., & Chen, X. (2025a). The use of generative AI by students with disabilities in higher education. The Internet and Higher Education, 66, 101014. https://doi.org/10.1016/j.iheduc.2025.101014 Zhao, X., Chen, X., Huang, V. H., Rollins, M., Carratù, M., & Shallari, I. (2025b). Students' Use and Attitudes Toward Generative Artificial Intelligence: A Comparative Study Between the UK and China. In Proceedings of the 58th Hawaii International Conference on System Sciences (pp. 4933-4940). https://doi.org/10.24251/HICSS.2025.594

AI Dialogue could serve as a bridge connecting key stakeholders, including EdTech developers, researchers, practitioners, users, and policymakers, within a shared platform. By bringing these groups together, it can enable the exchange of perspectives, support the identification of common themes, and highlight areas that require further attention. It can also help surface emerging solutions that are suitable for diverse educational, social, and cultural contexts. Importantly, AI Dialogue can help bridge top-down and bottom-up approaches to AI governance. While policymakers and institutions may provide regulatory direction, practitioners, users, and affected communities can contribute grounded insights from lived experience and everyday practice. This helps ensure that AI governance is not only shaped by high-level policy agendas, but also informed by the needs, concerns, and aspirations of those directly affected by AI systems. In this way, AI Dialogue can promote more inclusive and responsive international cooperation by ensuring that all voices are heard, particularly those of disadvantaged and underrepresented groups.

Current initiatives often approach AI governance primarily through a prevention-oriented model, focusing on risk management, regulation, safety, and compliance. While these are essential, the AI Dialogue should also strengthen an educational model of AI governance—one that prepares future generations with the awareness, knowledge, skills, and critical capacity to engage with AI governance in practice. In this regard, the AI Dialogue should build on existing educational initiatives, particularly the UNESCO AI competency frameworks, which have been widely used by educational institutions to support AI-related teaching, training, and capacity building. These frameworks provide an important foundation for embedding AI literacy, ethical awareness, responsible use, and governance-related competencies into education systems. The AI Dialogue could add value by engaging more actively with educational institutions, especially higher education institutions around the world. Universities and colleges are not only users of AI technologies; they are also key sites for research, innovation, professional training, and public debate. Most importantly, they educate future leaders, employers, policymakers, developers, teachers, and citizens who will continue to shape AI governance over time. By engaging higher education institutions globally into the discussion, the AI Dialogue could help ensure that AI governance is not treated only as a matter for policymakers and technical experts, but as a shared educational and societal responsibility. This would support a more sustainable approach to international cooperation by building long-term capacity, encouraging cross-cultural dialogue, and empowering diverse learners to participate meaningfully in shaping responsible AI futures.

I would recommend that, after collating the survey responses, the AI Dialogue should identify key themes, such as urgent questions to be addressed, emerging challenges, promising solutions, and areas requiring further collaboration. By analysing contributors' profiles alongside their responses, it would be possible to match stakeholders who share common interests, expertise, or concerns. Following this process, a structured network could be established, with thematic sub-groups formed around the identified priorities. These sub-groups would allow more focused and sustained conversations among stakeholders with shared interests, including policymakers, researchers, EdTech developers, educators, students, practitioners, civil society organisations, and representatives of disadvantaged groups. To support meaningful outcomes, funding could be allocated to these thematic networks to organise online meetings, workshops, and collaborative events. These activities could help stakeholders develop actionable projects that respond directly to the urgent questions and needs identified through the survey. In this way, the AI Dialogue would move beyond consultation towards sustained collaboration. It could create a community of key stakeholders who are actively engaged in AI governance and who can bring their own diverse networks, knowledge, and resources into the process. This would help ensure that the Dialogue is not a one-off exercise, but an ongoing mechanism for building relationships, sharing perspectives, developing practical solutions, and strengthening international cooperation on AI governance.

Our empirical research, including a global survey, shows that current discussions on AI and AI governance still lack perspectives from marginalised communities. These include people in Global South contexts, where participation may be limited by unstable digital infrastructure, unequal access to premium AI tools, and language barriers for those who do not speak English as a first language (Zhao et al., 2026a). A further survey with over 120 students with disabilities also shows that many students want their voices to be heard and wish to be involved in policymaking on AI and AI governance. However, their participation is often limited by structural barriers, including a lack of accessible channels for engagement and concerns about being accused of inappropriate AI use (Zhao et al., 2026b; 2025). To address this, the AI Dialogue should create more inclusive and accessible routes for participation. This could include multilingual consultations, low-bandwidth participation options, funded access to events, accessible online platforms, and dedicated forums for disabled students and other underrepresented groups. It is also important to create safe spaces where learners and marginalised communities can discuss AI use openly without fear of judgement, punishment, or surveillance. These communities should not be treated only as consultation participants. They should be actively involved in shaping agendas, identifying urgent governance questions, and co-developing practical solutions. By including those who are most likely to experience exclusion, bias, or unequal access, the AI Dialogue can become more socially grounded, equitable, and responsive to diverse global contexts. References Zhao, X., Miao, F., Xie, H., & Chen, X. (2026a). Exploring Student and Educator Challenges in AI Competency Development: A Comparative Analysis. Multimodal Technologies and Interaction, 10(3), 27. https://doi.org/10.3390/mti10030027 Zhao, X., Chen, X., & Cox, A. (2026b). Exploring the affordances of generative AI in academic writing for students with disabilities: A bottom-up approach to inform GenAI policies. Policy Futures in Education, 24(1), 59-81. https://doi.org/10.1177/1478210325139 Zhao, X., Cox, A., & Chen, X. (2025). The use of generative AI by students with disabilities in higher education. The Internet and Higher Education, 66, 101014. https://doi.org/10.1016/j.iheduc.2025.101014

At the University of Manchester (2026), we are hosting the first Inclusive AI Conference, which aims to bring key stakeholders into dialogue through plenary panels and evidence-based discussions. The conference includes stakeholders from organisations such as Microsoft, Adobe, UNESCO, Jisc, Wonkhe, and higher education leadership, alongside empirical research presentations and practical case studies. This provides one possible model for how the AI Dialogue could move from broad consultation to structured, action-oriented engagement. I would suggest that, after collating survey responses, the AI Dialogue could identify key themes and organise themed conferences, workshops, or online events around these priorities. These events could bring together relevant stakeholders as panellists, including policymakers, technology companies, researchers, educators, students, civil society organisations, and representatives from marginalised communities. In addition to panel discussions, the AI Dialogue could establish hackathon-style events based on the urgent questions and challenges generated through the survey responses. These could support collaborative problem-solving and help participants develop practical outputs, such as policy recommendations, educational resources, accessibility guidelines, institutional toolkits, or prototype solutions. This approach would ensure that the AI Dialogue is not limited to collecting views, but becomes a mechanism for turning evidence into action. By connecting empirical findings, stakeholder dialogue, themed events, and collaborative project development, the AI Dialogue could build a sustained international community committed to inclusive, educationally grounded, and context-sensitive AI governance. References The University of Manchester (2026). Call for Interest opens for 2026 Conference on Artificial Intelligence in Higher Education. https://www.manchester.ac.uk/about/news/call-for-interest-opens-for-2026-conference-on-artificial-intelligence-in-higher-education/

This is similar to Q17. So, I've used the same answer. At the University of Manchester (2026), we are hosting the first Inclusive AI Conference, which aims to bring key stakeholders into dialogue through plenary panels and evidence-based discussions. The conference includes stakeholders from organisations such as Microsoft, Adobe, UNESCO, Jisc, Wonkhe, and higher education leadership, alongside empirical research presentations and practical case studies. This provides one possible model for how the AI Dialogue could move from broad consultation to structured, action-oriented engagement. I would suggest that, after collating survey responses, the AI Dialogue could identify key themes and organise themed conferences, workshops, or online events around these priorities. These events could bring together relevant stakeholders as panellists, including policymakers, technology companies, researchers, educators, students, civil society organisations, and representatives from marginalised communities. In addition to panel discussions, the AI Dialogue could establish hackathon-style events based on the urgent questions and challenges generated through the survey responses. These could support collaborative problem-solving and help participants develop practical outputs, such as policy recommendations, educational resources, accessibility guidelines, institutional toolkits, or prototype solutions. This approach would ensure that the AI Dialogue is not limited to collecting views, but becomes a mechanism for turning evidence into action. By connecting empirical findings, stakeholder dialogue, themed events, and collaborative project development, the AI Dialogue could build a sustained international community committed to inclusive, educationally grounded, and context-sensitive AI governance. References The University of Manchester (2026). Call for Interest opens for 2026 Conference on Artificial Intelligence in Higher Education. https://www.manchester.ac.uk/about/news/call-for-interest-opens-for-2026-conference-on-artificial-intelligence-in-higher-education/