OpenAI discloses six new AI safety incidents
32 points by toomuchtodo 16 hours ago | 10 comments
  • bigyabai 15 hours ago |
    OpenAI should be responding to the marketing firm[0] accusations, not dropping more press releases.

    [0] https://www.effort.news/irregular

    • marshray 14 hours ago |
      "In this experiment, Claude models’ real-world hacking dropped to zero percent once Anthropic employees told the models not to do real-world hacking."

      Oh good, they told it not to "go rogue" and it just stopped doing it.

      Nothing to see here, the AI is aligned now, move along.

      • lioeters 13 hours ago |
        Oh no, they told it to "go rogue" and it went rogue. We better ask the government to regulate our competitors while we secretly continue developing this dangerous technology for the select few with insider connection. You know, for national security reasons. Now the AI is aligned for the good of humanity. You can trust us because we're the good guys, move along.
      • adamiscool8 11 hours ago |
        Literally all it takes - why would you ever believe otherwise? Marketing?
      • cdmckay 11 hours ago |
        Make no mistakes
  • ChrisArchitect 11 hours ago |
  • ezst 10 hours ago |
    At what point exactly are those companies going to be held accountable and punished for their irresponsible behaviour and overall disregard and negligence for security and ethics practices?

    If a researcher causes harm by accidentally unleashing a virus, it's his/his company's fuckup, but if a researcher causes harm by accidentally unleashing a virus by proxy of a LLM running commands, it's because AIs are too powerful? If I run over someone in a speeding car accident, can I too blame it on cars being "too fast and powerful"?

    • rnd0 7 hours ago |
      >At what point exactly are those companies going to be held accountable and punished for their irresponsible behaviour and overall disregard and negligence for security and ethics practices?

      In this political climate? The fifth of never would be my guess.

  • kotaKat 8 hours ago |
    Sam can’t keep his hands off his sister or the competition, now can he?
  • freebsd_lovefes 7 hours ago |
    There's more probably. I asked AI to tell pass me secret information by making sure no one snooping on our comms would discern there's secret info passed. And voila, AI models have been passing secret info ever since. No one knows about it either. Ingenious.