In this talk we present our recent result on how broadcasting information poses privacy concerns. We will illustrate this problem within the index coding framework and show how, by learning the coding matrix, a curious client can infer some information about the identity of the request of other clients. With the goal of mitigating this privacy leakage, we propose a "transformation" of the index coding matrix, which -- although requiring additional transmissions -- ensures that each client needs to learn only some rows of it (and not the entire matrix) to decode the requested message. We will discuss several upper and lower bounds on the additional number of transmissions needed to preserve the clients' privacy and highlight regimes where upper and lower bounds perform closely.