OpenAI launched GPT-Live on Wednesday, a new set of voice models that can listen and speak simultaneously in real time, responding with conversational fillers like "mhm" or "I understand" in 9 different voices. The model can delegate complex tasks such as web searches to background systems while continuing the conversation without interrupting the flow. OpenAI tested GPT-Live for safety risks including "self-harm, psychosis and mania, emotional dependence on AI, violence, and sexual content" after facing lawsuits over chatbot harms, including a recent California plaintiff alleging the chatbot exacerbated his bipolar disorder. The system rolls out globally in two versions: GPT-Live-1 for paid users and GPT-Live-1 mini for free ChatGPT users.