OpenAI has officially announced 'ChatGPT Health', a new dedicated feature designed to provide users with a specialized environment for health-related discussions. This initiative comes as OpenAI observes a significant existing demand, with over 230 million individuals utilizing the platform weekly to inquire about health and wellness topics.

The core functionality of ChatGPT Health involves segregating these sensitive conversations from a user's standard chats. This 'siloing' mechanism ensures that the context of personal health information remains confined to the Health section, preventing it from inadvertently surfacing in other, general interactions with ChatGPT. Should a user initiate a health-related query outside this dedicated section, the AI is programmed to prompt them to transition to ChatGPT Health for appropriate context management. Interestingly, while health conversations are siloed, the AI in the Health section can still reference information discussed in the standard experience. For example, if a user previously sought assistance for a marathon training plan, ChatGPT Health would retain this knowledge, understanding the user's fitness goals when discussing health matters.

A key enhancement of ChatGPT Health is its ability to integrate with personal health data from various wellness applications, including prominent platforms like Apple Health, Function, and MyFitnessPal. OpenAI has explicitly stated that data from these health conversations will not be utilized to train its underlying models, addressing potential privacy concerns. Fidji Simo, CEO of Applications at OpenAI, articulated in a blog post that ChatGPT Health is envisioned as a tool to address prevalent issues within the healthcare sector, such as high costs, accessibility barriers, overburdened medical professionals, and a notable lack of continuity in patient care.

However, the introduction of AI chatbots for medical advice also ushers in a new set of challenges and concerns. Large Language Models (LLMs), like ChatGPT, function by predicting the most probable response to a prompt rather than necessarily the most factually correct one, given their lack of inherent understanding of truth. This operational characteristic makes LLMs susceptible to 'hallucinations,' where they generate plausible but incorrect information. Reinforcing these limitations, OpenAI's own terms of service clearly state that its platform is "not intended for use in the diagnosis or treatment of any health condition." The company anticipates rolling out ChatGPT Health to users in the coming weeks.