For the AtProto apologists:
Literally every post transmitted over AtProto is now being fed into an LLM.
It is literally impossible to opt-out, because the protocol is designed to make safety impossible.
https://techcrunch.com/2026/03/28/bluesky-leans-into-ai-with-attie-an-app-for-building-custom-feeds/
This applies double to the AtProto apologists on Fedi.
You go off about any criticisms of the protocol, because maybe some group somewhere will figure out how to use it for their own empowerment.
But then you're silent about the design choices and AI tools that make that hoped-for future all but impossible.
BlackSky allows local-only posts to not get fed into this AI, because they're not actually transmitted with AtProto.
That's as good of a fix as I think is possible, given the protocol.
https://bsky.app/profile/rude1.blacksky.team/post/3miclqt2nb22z
Somebody who actually knows the protocol better can tell me if Ruby's solution will hold up if a BlackSky user uses Attie.
I'm not sure if this holds up if BlackSky users log into Attie, though. It might be like Zeitgeist where it just takes a single untrustworthy local user to expose everything.
If Attie only works directly with Relays than BlackSky should be safe.
If Attie allows users to authenticate access to the BlackSky AppView than BlackSky users obviously would not be safe.
The Fedilore Otter 🦦
