Microsoft is warning that attackers can plant hidden memory instructions inside AI systems to subtly influence recommendations and decisions. The company says this emerging threat could affect everything from chatbots to automated tools
Microsoft flags new threat: hidden prompts that skew AI
