What? MONITOR is a command like all other commands - you can run it from a client, or really, anything that can establish a TCP a connection to the Redis server. (Eg telnet.).
In your specific example, your client is also probably the one issuing the commands to store and retrieve the private messages in Redis - meaning, it already has access to all the data required for a breach. If you have an unrelated workload that is accessing a shared redis, but it does not need access to the sensitive data, perhaps, use an ACL to limit clients from the alternate workload to a different redis DB (which provides some level of isolation), or simply use a different redis entirely, and don't allow clients running the alternate workload to connect to the database that contains sensitive data.