AI is increasingly prevalent in our daily lives, and this trend is unlikely to change anytime soon. This comes with risks, but by understanding these risks, we can build AI systems that mitigate them.