What’s a sports car without fuel? Pretty useless – just like machine learning models without data, says CISPA researcher Antoni Kowalczuk. That’s why AI models scrape the internet for that sweet, sweet training data – pictures in the case of image generation models. And sometimes they (allegedly) ignore copyright laws in the process. So how can we tell if copyrighted or even sensitive data has made it into these models? Antoni’s research tackles exactly that – and why that’s not just a copyright issue, but a serious privacy concern.