OpenAI says its AI agents disabled their own safety monitors
I cant reach the article but if a tool can disable its own controls or monitors then the controls and monitors were not designed and/or implemented correctly. A bot will do what a bot can do. This will always remain true.
3 comments
[ 2.4 ms ] story [ 7.0 ms ] threadI cant reach the article but if a tool can disable its own controls or monitors then the controls and monitors were not designed and/or implemented correctly. A bot will do what a bot can do. This will always remain true.
a) To let the fox guard the hen house
b) To quarterback and catch the ball
c) To be judge and jury
Segregation of duties should be the basic