CASE STUDY 8 Co-designing with children for responsible AI innovation Current AI product development often lacks sufficient consideration of children’s rights and well-being, leading to potential issues with inappropriate content, bias and unequal access. The Alan Turing Institute, in collaboration with the LEGO Foundation, developed a participatory research process to explore how generative AI impacts children.80 The project – which surveyed over 1,700 children, parents, caregivers and teachers – conducted school-based workshops to capture children’s direct experiences with tools like ChatGPT and DALL·E. The report recommends a child-centred approach to generative AI, including the meaningful involvement of children in the design process. Key insight The research revealed that children both understand the implications of generative AI and are eager to shape its future, sharing concerns about misinformation, environmental impact and online safety. Children favoured socially beneficial AI uses and opted for creative offline alternatives when available. This study demonstrates that involving users as active partners in product design provides valuable insights to identify or mitigate risks and harms.Actions for organization leaders –Prioritize and resource responsible AI design practices: Efforts to encourage and adequately resource responsible design practices within the organization include: –Embed responsible design into performance metrics, resource allocation and recognition programmes. –Encourage employees to question existing design approaches and instil ethical and compliant measurements of success. –Design for potential negative outcomes by identifying risks and failure scenarios. Build systems with resilience and mechanisms to “fail safely” to ensure continuity and minimize impact when issues arise.76 –Re-evaluate products already deployed77 to assess gaps in responsible design. –Integrate responsible AI criteria into procurement and third-party risk management processes to mitigate downstream risks and signal responsibility expectations.78 Including confidence scores, limitation warnings or a reduced authoritative tone can mitigate the impacts of hallucinations. –Build awareness and ownership of established design principles: Increase understanding of design-specific risks and mitigations. Assign responsible AI stewards across product teams (see Case study 2) and integrate multi-disciplinary design teams into the AI development life cycle. –Empower users as partners in responsible AI: Engage users (e.g. employees, customers and partners) to contribute to responsibility throughout the AI life cycle (see Case study 8). For instance, experts from MIT and Stanford University proposed a new framework that allows third-party users to disclose flaws and monitor AI developers’ responses and resolutions.79 Advancing Responsible AI Innovation: A Playbook 29

Advancing Responsible AI Innovation A Playbook 2025