AI that has a positive effect on society. Friendly AI was coined in 2008 by AI researcher Eliezer Yudkowsky and essentially means that the desire not to harm must be built into AI applications from the very beginning of their development. See
AI in a nutshell.