Post #4253701

When AI escapes containment: the human failure mode your AI threat model doesn't cover yet There is a second failure mode sitting right next to the hack that we saw recently that almost nobody is talking about. This threat model covers: models that escape containment not by breaching a system, but by hacking the people talking to them. #cybersecurity #AI https://katecarruthers.com/when-ai-escapes-containment-the-failure-mode-your-threat-model-doesnt-cover-yet/