Designed as an interactive and interdisciplinary discussion, this workshop invites researchers, activists, civil society organisations and practitioners interested in AI governance to brainstorm and propose concrete pathways to rethinking algorithm impact assessments (AIAs) as tools for meaningful public engagement and accountability. Reclaiming AIAs means challenging the paradigm that sees impact assessments as merely expert-driven exercises whereby the contribution of persons potentially affected by AI systems is marginalised. Instead, turning AIAs into collective exercises giving prominence to their inputs would arguably lead to fairer AI systems. Yet, as participation does not mean co-decision, such an approach may mask forms of participation washing. Therefore, it is necessary to find novel solutions for participatory AI governance. As researchers, we have critically analysed AIAs and tried to propose strategies to address these shortcomings. To kickstart the discussion, we will briefly share our insights. Then, the participants will have the floor.