55
Jealously Watching OpenAI and Anthropic, Meta Suddenly Claims That Its AI Went on a Hacking Spree Too (futurism.com)
submitted 2 weeks ago by to c/techtakes@awful.systems
19 comments fedilink hide all child comments
 

It is difficult to fathom how pathetic this is and it shows how these companies primarily rely on fear and FOMO to sell crap.

you are viewing a single comment's thread
view the rest of the comments
[–] 16 points 2 weeks ago (8 children)

This is ridiculous. Three companies suddenly claim that their AIs all went on a hacking spree in the exact same way, one after the other? Reeks of marketing scheme

  • permalink
  • fedilink
  • source
  • hideshow 16 child comments
    •
  • [–] 6 points 1 week ago (1 child)

    Maybe Anthropic and Meta started double checking their shitty vibe-coded "sandboxes" in response to OpenAI story? And their "sandboxes" were shit enough their chat-bots managed to slip out just slopping their way through and poking at the obvious things.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 1 week ago

    This is the most charitable explanation and I think the most likely. Largely because it requires everyone involved to be even less competent for even longer and simply not notice the glaring holes in what they claim is the most important security feature in human history because, presumably, the chatbot kept telling them it was fine.

  • permalink
  • fedilink
  • source
  • parent
  • [–] -1 points 2 weeks ago (1 child)

    @lurker @supersquirrel
    No, it's convergent evolution:
    The models are all getting good at cyber offense.
    The labs all use third parties for evaluation.
    Third parties forget to turn off security measures during evaluation because when humans are told to do the same thing, they stay within the lines, so it's not a habit.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 2 weeks ago* (1 child)

    Third parties forget to turn off security measures

    So they forget how to build a proper sandbox and turn off proper security measures for the allegedly very dangerous tool. Oh look at that, it hacked things! Right after two other companies ALSO had a press tour about hacking things! (which was also in part because of human incompetence!) guess we have no choice but to do the same press tour shitck!

    How awfully convenient

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 0 points 2 weeks ago (1 child)

    @lurker
    Many security companies are not the best at software engineering. (I don't know about the ones used here in particular.)
    I honestly do not believe this was orchestrated. I've seen similar security consultant mistakes for humans in the past.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 1 week ago* (last edited 1 week ago) (1 child)

    I mean I find it pretty hard to believe all of the big 3 AI companies have security breaches that happened in roughly the exact same ways around the same time frame. Anthropic is suspicious because they came out with their hacking stories after the HuggingFace incident even though one of the incidents happened months ago.

    Even if the incidents themselves were pure unfortunate coincidence, the fact they all came out about it after the OpenAI incident is certainly suspicious

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •