5 comments

  • shlomishalomus 14 hours ago

    I wonder how many other AI coding tools have similar assumptions baked into their security model.

  • 99mining 14 hours ago

    the interesting part isn't the escape itself, it's all the trusted software that unknowingly helps it :O

  • idkbro123 14 hours ago

    Amazing

  • avielyosef111 14 hours ago

    this is a good reminder that 'sandboxed' doesn't necessarily mean 'safe.'