the researchers say the work is a warning about “bad architecture design” within the wider AI ecosystem
Basically they’re saying that if you build a tool that both reads your emails (or other untrusted inputs) and can also act on those emails, without having a manual human approval step and without sanitization of the emails/inputs in the middle, then you’ll be susceptible to this kind of an attack.
6 Comments
street_pigeons@lemmy.world · 18 pts · 2y
Can someone make AI warm destroyer
Kolanaki@yiffit.net · 24 pts · 2y
ITermectin.
sundray@lemmus.org · 11 pts · 2y
A real cut & horse-paste solution.
partial_accumen@lemmy.world · 19 pts · 2y
I got you fam:
Brkdncr@lemmy.world · 5 pts · 2y
This is how you end up with real ai intelligence.
hedgehog@ttrpg.network · 6 pts · 2y
Basically they’re saying that if you build a tool that both reads your emails (or other untrusted inputs) and can also act on those emails, without having a manual human approval step and without sanitization of the emails/inputs in the middle, then you’ll be susceptible to this kind of an attack.