Artificial intelligence has changed surveillance from observation to inference, with systems able to decode emotions, intentions, and beliefs based on behavioral and biometric measurements. The article proposes the concept of cognitive privacy as the right to mental autonomy and to avoid algorithmic manipulation. It explores the inability of current constitutional and statutory structures that were developed in an analog world to deal with inferential harm. It finds legal and normative loopholes in data collection, inference, and governance through the analysis of AI-driven surveillance architecture. To overcome these shortcomings, this paper suggests a four-layered cognitive-privacy protection model that includes sensor boundaries, inference control, interface transparency, and institutional oversight. It concludes that the only way to protect cognitive liberty is to redefine privacy as a control over inference as opposed to secrecy of information that is an indispensable measure towards the maintenance of democratic agency in the era of intelligent surveillance.