I don’t have an answer, but a slightly different perspective. Many different segments have a deep interest in using highly secure encrypted communications: politicians working on deals within/between governments (that should be auditable, but many try to avoid that), whistleblowers, organizers operating in adverse governments, dissidents, terrorists, pedophiles with a lot to lose (similar to Epstein’s network), healthcare professionals trying to talk to patients or other doctors in a hippa world, illegal transaction networks, attorneys with clients, VCs trying to debate the future of the world, companies trying to preserve trade secrets, you name it. It takes one of the egregious bad actors using the system to commit a crime worthy of public attention before the entire system is justifiably unpacked, banned, or considered a signal of bad intentions.

How can a system be made decentralized, but able to self-police against legitimately, publicly agreed upon bad behavior? If the system is able agree upon and exclude legitimately bad behavior automatically, the governments would not have a claim upon needing to police it and regular users would probably find it beneficial as well.

How could the self policing possibly happen?

Maybe you have a blockchain of anonymized encrypted messages that is read by open source scanning bots - if enough independent bots flag a message, then a group of anonymous judges can adjudicate to ban those user accounts?

Encryption is one challenge, but if you want true ubiquitous privacy, you need to deliver internal safety to prevent the need for external policing of activity. Social creatures of any species from dolphins to macaques have evolved some kind of internal behavior policing mechanism or trust is lost, and as such the system of value exchange grinds to a halt.