ChatGPT health and AI for healthcare: From commercial personal assistant to public health infrastructure

The launch of OpenAI’s ChatGPT Health for individuals and ChatGPT for Healthcare for institutions marks a major expansion of large language models into health and healthcare delivery. While these products promise efficiency gains and improved access to information, they also raise significant concerns for privacy, patient safety, and the governance of digital health infrastructures. This paper provides an evaluation of OpenAI’s health-focused products by examining their systemic implications and outlining public-interest governance models that could better align generative artificial intelligence (AI) with health system goals. Using a document-based analysis of publicly available vendor communications, technical documentation, and policy pages—interpreted through sociotechnical frameworks in health informatics—we evaluate product architectures, data flows, validation claims, and governance mechanisms. Vendor materials indicate that both products may facilitate large-scale aggregation of sensitive health and behavioral data, enabling longitudinal health profiles that raise unresolved questions about data retention, secondary use, reidentification, and cross-border transfers. Despite public claims of safety and high benchmark performance, available validation appears concentrated on narrow tasks and proprietary internal assessments, which cannot substitute for appropriate clinical governance or comprehensive regulatory evaluation. These developments may also deepen dependence on a single global vendor, creating structural risks for digital sovereignty and health equity, particularly in jurisdictions with weaker data protections. Ultimately, generative AI in health should not be framed primarily as a multi-purpose consumer application operated by a monolithic platform, but as a specialized tool embedded within publicly governed digital health infrastructures.