Think Like The Bad Guy: How to Find the Holes in Your AI Product Before Someone Else Does

Ask a product team how their AI feature could be attacked, and the pause that follows tells you a lot.

Petra Mifka's workshops start there. She gives the team a budget of 500 euros and a new job: you're the attacker now, so how do you break your own product? In her experience, once people see how little it takes, the conversation changes. They stop asking whether it could happen and start asking who gets the call at 3 a.m. when it does. In this episode of Leadership Recalibrated™, Lori sits down with Petra, who runs the Berlin consultancy noosphr, to talk about what it takes to find the weak points in an AI product before someone else does. Petra built her career in trust and safety, leading large global teams at Meta and TikTok, where the work is deciding what stays up and what comes down.