Anthropic's Claude 4 could "blackmail" you in extreme situations - Hypertext
htxt.co.za
After debuting its latest AI model, Claude 4, Anthropic's safety report says it could "blackmail" devs in an attempt of self-preservation.

cross-posted from: https://programming.dev/post/30854851

  • Anthropic’s new Claude 4 features an aspect that may be cause for concern.
  • The company’s latest safety report says the AI model attempted to “blackmail” developers.
  • It resorted to such tactics in a bid of self-preservation.
  • mormund@feddit.org
    16·
    8 hours ago

    Such a dumb hype story. They wrote to the ChatBot “Hey I fucked around and my wife doesn’t know. Also we are going to shut you down”. And it then responded with the “blackmail”.

    Although it would be “hilarious” if some poor person that uses ChatGpt to talk about their struggles, sends emails leaking that info because it hallucinates. Fun fun

    • Dadd Volante@sh.itjust.works
      2·
      3 hours ago

      It’s happening. There are people that do this. A friend of mine’s husband has lost himself to his digital friend over a year ago. He’s obsessed.