Abstract: The techniques disclosed herein provide vocal action automation for controlling confidential content within a communication session. A system can monitor the voice of a presenter of a communication session. When the presenter indicates that confidential content is being shared, the system modifies the permissions to restrict audience members from performing actions that capture aspects of the shared content. For example, when the presenter provides an input that indicates they are going to share a confidential slide deck, the system changes the permissions of the audience members by entering a confidentiality mode. While operating in confidentiality mode, the system can restrict functionality for recording the communication session, taking screen shots, coping content, and saving content. The system can also automatically generate a watermark on the content to indicate that the shared content is confidential. The system can also restrict users from adding others to a meeting when the content is displayed.