Techniques for polluting phishing campaign responses with content that includes fake sensitive information of a type that is being sought in phishing messages. Embodiments disclosed herein identify phishing messages that are designed to fraudulently obtain sensitive information. Rather than simply quarantining these phishing messages from users' accounts to prevent users from providing “real” sensitive information, embodiments disclosed herein analyze these phishing messages to determine what type(s) of information is being sought and then respond to these phishing messages with “fake” sensitive information of these type(s). For example, if a phishing message is seeking sensitive credit card and/or banking account information, some fake information of this type(s) may be generated and sent in response to the phishing message. In various implementations, a natural language processing (NLP) model may be used to analyze the phishing message and/or generate a response thereto.