A robot for detecting a dangerous situation using artificial intelligence includes a memory configured to store a voice recognition model for inferring whether a current situation is the dangerous situation from voice data and an image recognition model for inferring whether the current situation is the dangerous situation from image data, and a processor configured to acquire one or more of the voice data or the image data and output a notification indicating the dangerous situation when the dangerous situation is detected from the voice data using the voice recognition model or when the dangerous situation is detected from the image data using the image recognition model.