A Saudi national has been charged with secretly planting a recording device in one of OpenAI’s offices in California, federal prosecutors announced.

Abdulrahman Alfozan, who worked at the AI firm as an engineer, was charged federally with using or endeavouring to use an electronic, mechanical, or other device to intercept oral communications.

OpenAI said it had discovered a “sophisticated recording device” hidden in their San Francisco office and alerted the FBI. The company said it used its AI models to track that it had been planted by an employee.

  • egregiousRac@piefed.social
    link
    fedilink
    English
    arrow-up
    72
    ·
    2 days ago

    The company said it used its Al models to track that it had been planted by an employee.

    Bullshit.

    • They had CCTV of him planting it which they could positively identify using badge records.
    • The device had a camera on it that captured his face.
    • Security searched an unattended bag elsewhere in the building and found a matching device and his wallet inside.
    • He used the company’s product to search for guides on how to plant it.

    None of that is using AI models to track who planted it.

    • orclev@lemmy.world
      link
      fedilink
      arrow-up
      34
      ·
      2 days ago

      They’ll never miss an opportunity to lie and claim their model can do more than it’s capable of. It should be obvious at this point that any time OpenAI (or most of the AI companies for that matter) claim their AI has done something you should be incredibly suspicious of those claims because it’s a near certainty they’re either highly exaggerated or just plain false.

      • tidderuuf@lemmy.world
        link
        fedilink
        arrow-up
        10
        ·
        2 days ago

        OpenAI: “Our coffee machine is ran by our AI models. Once we grind the beans, put the water in, place the filter, put the mug under, press the button… Our AI does amazing work behind the scenes to extract the coffee into our delicious drinks. Buy more billionaire welfare bonds, I mean AI stock!”

    • deegeese@sopuli.xyz
      link
      fedilink
      arrow-up
      5
      arrow-down
      1
      ·
      2 days ago

      Yeah they’re probably lying but it’s not really a stretch to imagine they’re dogfooding use cases like that.

      They already have a log of all badge swipes and a DB of meeting invites. Not hard to ask a LLM “who badged into this room in the past 2 days when they didn’t have a meeting here?”. That would narrow it down to the janitor plus Mr Bozo.

  • kreskin@lemmy.world
    link
    fedilink
    arrow-up
    3
    ·
    1 day ago

    Saudis were messing around inside twitter too, a long while back-- using it to hunt for dissidents. That government really is a bunch of evil little shits. We should not be military allies with them.