Researchers at Northeastern University recently found that OpenAI had implemented stronger safeguards around questions related to suicide and self-harm, but concluded that chatbot companies in general have failed to extend specific protections regarding other mental health conditions, including substance abuse and eating disorders.