By code or by command, cybercriminals are circumventing ethical and safety restrictions to use generative AI chatbots in the way that they want.