Mobile devices make it possible to create, store, access, share or publish personal content on the Internet, anywhere and at anytime. This leads to situations of potential intentional or unintentional misuse of content as well as privacy issues. Recent techniques involving the use of contextual information focus on access of documents stored in clouds, or authentication for secured Web sites. These techniques or more traditional solutions, such as steganography or Digital Rights Management, do not empower the user itself, or data controller in professional settings, with a fine-grained control of the access to or manipulations actions on documents stored on mobile devices, e.g., copying, sharing, etc. In this paper, we propose SmartContent, a novel approach for content protection and privacy. Documents are active and context-aware documents that sense and analyse their current context, e.g., location, noise, neighbouring devices, social network, expiration time, etc. Based on user provided policies, they grant, deny or limit access and manipulation actions, or destroy themselves if necessary. We present the generic model of SmartContent, a concrete architecture and an implementation of a proof-of-concept specifically designed for mobile devices. We deployed it on tablets and showed that a picture dynamically reveals or conceals itself based on sensed context or on changing policies. The implementation leverages the SAPERE middleware specifically developed for context-aware systems.