CreepyHunterX

New private messages are protected by end-to-end encryption (E2EE), but the ‘photo sharing’ in the name, isn’t. And A/SIDE stores the feed, so they can read it. The App Store label "User Content Linked to You" refers specifically to the feed, so there is no contradiction here. So the feature they named the product after is the one that isn't private.

Either the payload was already present in the captured snapshot (in which case the original is at fault), or it was added by the server or the rendering layer of the archive.today service

The real risk here isn't AI training - It's deanonymization. People posted in a "private" community under nicknames.

However, the server administrator enables the "discoverability" feature for everyone. So the entire history of members who never chose this option suddenly becomes discoverable and linked to their accounts. Use the same nickname on GitHub or X - and in a matter of seconds, it will point to the same person.

DoH allows this to be bypassed, as they operate over port 443.

However, DoT uses a separate port (853) that can easily be blocked entirely, whereas filtering DoH requires relying on SNI or lists of IP addresses.

What are you merging on - a shared PGP key or a crypto address or every hit it's own node? And what happens if different people happen to use the same nickname?

If Verisign frees up domain.name, the new owner could inherit email still directed to the old address, for example: password resets.

Does your application check DMARC alignment? Otherwise, a spoofed 'from' address will pass through. And also the sender's policy alone is not enough because, gmail.com and outlook.com have DMARC p=none.