Recommender systems can improve the user experience for users of all ages by surfacing relevant content. However, the design of algorithms can expose children to harmful content and to content that can cause harm to build when encountered repeatedly over time. This session will bring together stakeholders from industry, civil society and a regulator to discuss the risks that children face when encountering recommended content, in particular, risks of cumulative harm. Panelists will consider questions like ‘what is cumulative harm?’ and ‘how can the design of recommender algorithms mitigate risks of cumulative harm to children?’. Panelists will also consider what they see as the opportunities and challenges of taking a safety by design approach to recommender algorithms. Panelists will also reflect on how T&S interventions for recommended content differ from interventions for searched-for content and the extent to which the risk of cumulative harm is a factor in these decisions.