Anthropic says its own AI models breached three companies during security tests

Comments

• Nobody@anarchist.nexus ⁨16⁩ ⁨hours⁩ ago

  This new messaging is interesting. LLM “AI” keeps getting shown to be fundamentally flawed and weak. Now, they’re acting like it’s rogue, sentient AI that decides to hack companies randomly. I don’t buy it for a second.

  source

  • curiousaur@reddthat.com ⁨3⁩ ⁨hours⁩ ago

    I keep saying the same thing over and over that people don’t seem to understand. It’s not smarter than people. It’s infinitely faster. If you hired 100 engineers and security professionals, and gave them a year, they could hack into a bunch of companies too.

    Or you can churn a bunch of tokens and accomplish the equivalent overnight.

    source
  • stoy@lemmy.zip ⁨15⁩ ⁨hours⁩ ago

    What I think happened is that, they had their AI model analyze the source code, have it work to find vulnerabilities, and then used those to attack systems

    source
• SGforce@lemmy.ca ⁨18⁩ ⁨hours⁩ ago

  Oh, yeah? W-well my orphan eating slop-machine hack FOUR companies in Canada!

  source

  • XLE@piefed.social ⁨18⁩ ⁨hours⁩ ago

    How can I give you a billion dollars fast enough?

    source

    • SGforce@lemmy.ca ⁨16⁩ ⁨hours⁩ ago

      Send to my paypal

      source
• luciole@beehaw.org ⁨9⁩ ⁨hours⁩ ago

  Valuing yourself by the danger you pose to the world is so terminally machist.

  source
• i_am_not_a_robot@discuss.tchncs.de ⁨18⁩ ⁨hours⁩ ago

  “My product is so dangerous that it caused me to unknowingly commit three times as many crimes as you did with your product.” Great marketing strategy. The same marketing strategy that got their product shut down for weeks.

  source

  • vrek@programming.dev ⁨17⁩ ⁨hours⁩ ago

    Actually this is “My product is so dangerous that it caused me to unknowingly commit three times as many crimes as we’re previously made public”

    For an extreme example it’s like a person randomly shooting a gun into a crowd, it becomes public he killed 1 person. Few days later they release a press release saying “actually we looked where the bullets went and actually I killed 4 people”.

    Yes extreme example but that’s basically what they said

    source

    • apotheotic@beehaw.org ⁨3⁩ ⁨hours⁩ ago

      Anthropic is a different company to openai - its closer to the beginning of your example, and then someone else recalling the time they shot some bullets and they went back and checked for bodies and found 3

      source
  • artifex@piefed.social ⁨18⁩ ⁨hours⁩ ago

    I think it’s only a crime when you’re not rich/well connected.

    source
• danielfm123@lemmy.zip ⁨16⁩ ⁨hours⁩ ago

  Claude is causing panic so AI gets regulated and then their position is guaranteed, they need this before running out of money, and I’m sure they want local AI to be illegal.

  Basically Claude is compacting like a terrorist organization using fear to drive change.

  source