The company says ChatGPT is trained to direct people to suicide and crisis hotlines. OpenAI said some of its safeguards might not kick in during longer conversations and it is working on preventing that from happening. Read full story