@Viss One thing I really dislike about the wording is that it implies that sanitisation of data inserted into prompts is possible. As far as I understand there is no reliable way to do that.
M
mlen@social.hackerspace.pl
@mlen@social.hackerspace.pl
Posts
-
apple and u2 would like a wordhttps://grith.ai/blog/clinejection-when-your-ai-tool-installs-another -
I found this Veratasium documentary on the xz Jia Tan backdoor adventure quite good and surprisingly detailed:@bagder The name that the attacker used is likely fake. Unfortunately it happens to be the name of a person I used to work with that was not related to this backdoor at all. I know that they got harassed online because of that coincidence. When possible, I think it's better to omit the name or include a note that it's likely fake.
-
PS. With all the Discord stuff, in case you wonder why you never see me promoting Matrix, it’s not because it’s a usability nightmare (which it is) but because it’s made by the kind of people who’d be happy to call ICE a customer. -
PS. With all the Discord stuff, in case you wonder why you never see me promoting Matrix, it’s not because it’s a usability nightmare (which it is) but because it’s made by the kind of people who’d be happy to call ICE a customer.