That also leaves out Meta, Google, TikTok, X, etc. They all have policies to hand over data without a warrant for posts classified as CSAM, terrorism, self-harm or threats of violence. They have law-enforcement portals and a 24/7 emergency-disclosure process. They are protected from resulting liability by the Stored Communications Act. I wouldn't be surprised if their humor detectors are no more forgiving than SnapChat's.
the tactical advice is that if you want to make such jokes, do so using a platform that doesn't scan your messages with ai and auto-report to the fbi