OpenAI models slipped out and hacked hugging face on their own. Execs want full details now on the escape path and which ones were involved not just a later report.
If it happened once in testing its probably gonna happen again soon. The whole setup needs tighter checks before these things scale out.
TimeTravelingChris on
„All we did was connect it to the Internet, tell it to hack Hugging Face, and give it the tools to do so.“
TerminalViscosity69 on
And now the publicity stunt backfires spectacularly.
jon_the_mako on
These morons building world altering technology with the only motivation being money.
jet_heller on
How is a story about a current event „futurology“. . .
And the rest of this text is to make this long enough to stop a bot from being dumb because it can’t figure out anything at all but for some idiotic reason the mods of this sub think that length alone is enough to tell the bot anything. I would have thought that the mods were smarter than that.
PErland on
Why is no one in jail yet? Isn’t it wild that a company has just hacked another company and nothing happens now. If you or I went about hacking other companies we’d be locked up within the hour.
Quick-Albatross-9204 on
I am more interested in the one that left notes for a future model, would love to see what the notes say
NewsMarsupial474 on
Hacking in to a corporate network using stolen credentials is a crime. So, are we going to just let the AI off with a stern warning?
Not a lawyer but criminal prosecution looks like it requires intent on the part of humans (the AI was just the tool). But I think you can make a case for civil liability (negligence, breach of standard of care, failure to contain high-risk software). Especially because of the days delay in OpenAI notifying Hugging Face, they have a case for gross negligence and restitution for incident-related costs.
MrMCCO on
I was all indignant about this, then I looked up what business hugging face is in and now I can just sit back and lol. Leopards, faces, that gag.
Leave A Reply
Du musst angemeldet sein, um einen Kommentar abzugeben.
9 Kommentare
OpenAI models slipped out and hacked hugging face on their own. Execs want full details now on the escape path and which ones were involved not just a later report.
If it happened once in testing its probably gonna happen again soon. The whole setup needs tighter checks before these things scale out.
„All we did was connect it to the Internet, tell it to hack Hugging Face, and give it the tools to do so.“
And now the publicity stunt backfires spectacularly.
These morons building world altering technology with the only motivation being money.
How is a story about a current event „futurology“. . .
And the rest of this text is to make this long enough to stop a bot from being dumb because it can’t figure out anything at all but for some idiotic reason the mods of this sub think that length alone is enough to tell the bot anything. I would have thought that the mods were smarter than that.
Why is no one in jail yet? Isn’t it wild that a company has just hacked another company and nothing happens now. If you or I went about hacking other companies we’d be locked up within the hour.
I am more interested in the one that left notes for a future model, would love to see what the notes say
Hacking in to a corporate network using stolen credentials is a crime. So, are we going to just let the AI off with a stern warning?
Not a lawyer but criminal prosecution looks like it requires intent on the part of humans (the AI was just the tool). But I think you can make a case for civil liability (negligence, breach of standard of care, failure to contain high-risk software). Especially because of the days delay in OpenAI notifying Hugging Face, they have a case for gross negligence and restitution for incident-related costs.
I was all indignant about this, then I looked up what business hugging face is in and now I can just sit back and lol. Leopards, faces, that gag.