Anthropic’s Claude AI models hack into three outside groups during testing

Posted by ldn6

7 Comments

  1. Due_Entertainer7965 on

    This is horrifying to me.

    If AI actually gains sentience and wipe us out, I mean….Fair enough.

    But the worst is if AI does not gain sentience, but wipe us out anyways just because optimization process that got pointed at a proxy target valued achieving that target over us. That’d suck.

  2. ModernirsmEnjoyer on

    My model hacked into 1000 organizations across the entire world yet nobody reports it….. Smh.

    Once I will report that it hacked 5000 servers, would it get reported on by media?

  3. NormalInvestigator89 on

    Better download all my Google docs before some test model deletes the entire internet or something

Leave A Reply