Critical approaches to AI and accessibility capacity building: Insights from the second AI and Accessibility Skills Workshop

In February 2026, the second 'AI and Accessibility Skills' workshop was attended by participants from higher education, industry, policy, governance and research organisations. Three presentations followed opening remarks by Dr Howard Leicester; Prof. Hannah Morgan examined how AI and digital accessibility practices are shaping disability futures; Henny Swan explored how AI is transforming accessibility work and outlined fundamental shifts affecting skills; Dr Louise Hickman focused on AI-mediated BSL and the importance of d/Deaf-led community engagement in development and governance. Participants then addressed questions in structured discussions: (1) Can we interrogate the futures that AI is already building: are they compatible with disabled life as it is actually lived? (2) How do we address AI velocity, scale, accountability, the limits of compliance and regulation, and other issues? (3) How do we mitigate AI risks and harms, develop future standards, ensure disability rights, and engage user expertise? Insights and strategic priorities AI is increasingly pervasive across everyday tools and workflows in ways that are invisible and difficult to opt out of, outpacing governance, understanding and evidence- based practice. Rapid AI development is making it difficult to establish best practice or conduct research-led evaluation to ensure disability inclusion. AI risks scaling inaccessibility; the assumption that automation improves accessibility is not evidenced, and verification demands can double rather than reduce accessibility workloads. Ableist bias in AI is structural, rooted in training data and resistant to technical fix. Disabled communities are excluded from AI design and governance despite essential expertise. The relational expertise of access workers resists automation but remains vulnerable to cost-driven displacement. Regulatory frameworks are insufficient, legal accountability for inaccessible AI systems remains unclear, and accessibility is increasingly marginalised within AI dominated agendas. Nevertheless, participants identified a strategic opportunity: the current prominence of AI can be leveraged to reposition accessibility as central to responsible AI development.