Governments are buying AI to summarise citizen input — often without seeing inside the black box. This working group is building the shield: a checklist any public official can run before procuring or deploying a digital participation or AI system. This is our second working session, and you don't need to have been at the first one to take part. It's a working conversation, not a presentation — the kind of session where the value comes from putting different angles in the same room: someone who procures tools, someone who builds them, someone who designs the participation they're meant to serve. We'll share the current version of the Civic AI Checker — a democratic-quality layer that sits on top of the EU Model Contractual Clauses for AI, asking the question those clauses don't: does this system strengthen deliberation, or quietly automate it away? Then we open the floor: where does it hold up, where does it fall short, and what would make it genuinely useful in your context. From there we start the Clause Bank — the contractual clauses and technical standards that make transparency, contestability and inclusion enforceable, not just stated as intentions. Bring whatever you have: a red flag you've seen in a real deployment, a clause that worked, or simply a question from your own field. All of it feeds the checklist. For procurement officers, digital transformation directors and participation designers — and the researchers and civil-society partners working alongside them. Newcomers welcome.