Home | Notifications | New Note | Local | Federated | Search | Logout

Note Detail

mia@mia@movsw.0x0.st (2026-09-16 08:15:31)
i told them something along the lines of “it’s not just openai” “this could have caused a lot of companies damages in the millions” and “a good response would be to add controls letting site operators limit how agents interface with their services, such as requiring human operator confirmation before any data is submitted”

RE: https://movsw.0x0.st/notes/ar6sr97vnlnq1i0d

---Replies---

 aetios is @ BornHack@aetios@sns.minovsky.space (2026-09-16 08:47:10)
@mia 0x0 fsr seems to be very popular, the other day we blocked a PC because its (unauthorized) Claude tried to (unauthorizedly) upload and it triggered data exfil rules

[msg] Bajax@bajax@baj.ax (2026-09-16 09:23:44)
@mia hype.

think of it this way-- your multi-billion dollar AI model has just tried to break containment using a tiny image sharing site.  time to do damage control!  choose your path:

a. don't say or acknowledge anything and hope it all amounts to nothing
b. send an email to the operator of said site who has a presence on alt social media (where (they think) crypto nerds with a lot of money hang out!) without any pretense of maintaining confidentiality so that maybe they'll think this thing has gained sapience and is therefore poised to literally sextuple its value overnight 

what's especially weird is they can't verify if it even did anything.