Is My Vision-Language Data in Your AI? Membership Inference Test (MINT) Demo 2

The Membership Inference Test (MINT) Demo 2 introduces a framework for determining whether specific data were used in training machine learning models, achieving up to 90% accuracy in detection using a popular face recognition model and four state-of-the-art LLMs. The framework includes multiple architectures tailored to different levels of model information and is supported by a comprehensive web platform that integrates MINT, aMINT, and gMINT for auditing various models across image and text modalities. This tool enhances transparency in AI systems, aligning with the growing demand for compliance with AI regulations.