OpenAI rolls out AI suite for hospitals and healthcare systems

OpenAI has introduced a set of healthcare-focused products intended to support clinical, administrative and research work while meeting U.S. health data privacy requirements.

The offering, called OpenAI for Healthcare, includes ChatGPT for Healthcare, which is now becoming available to healthcare organisations, and the OpenAI API, which is already used across parts of the healthcare technology sector. The AI parent said the products are designed to support organisations’ compliance with the Health Insurance Portability and Accountability Act, or HIPAA.

It is being rolled out to healthcare systems and academic medical centres, including AdventHealth, Baylor Scott & White Health, Boston Children’s Hospital, Cedars-Sinai Medical Center, HCA Healthcare, Memorial Sloan Kettering Cancer Center, Stanford Medicine Children’s Health and the University of California, San Francisco, the company said.

The company said healthcare providers face growing demand, rising administrative workloads and fragmented medical information, while adoption of artificial intelligence is increasing unevenly across the sector. According to the American Medical Association, physicians’ use of AI nearly doubled in one year, though many clinicians still rely on personal tools due to slower adoption by institutions operating in regulated environments.

OpenAI for Healthcare is intended to provide organisations with a centralised, enterprise-grade AI foundation so clinicians and staff can use shared tools while supporting regulatory requirements, as the company said.

The programme is said to be designed to support clinical reasoning and reduce administrative work. Features include AI models tailored to healthcare workflows, evidence retrieval with citations from peer-reviewed research and clinical guidelines, and integration with institutional systems so responses can reflect an organisation’s policies and care pathways.

The product also includes reusable templates for tasks such as discharge summaries, patient instructions and clinical correspondence, along with role-based access controls and centralised user management. The patient data remains under the organisation’s control, with options such as audit logs, data residency, customer-managed encryption keys and a Business Associate Agreement. Content shared with this IP is not used to train the chatbot’s models, the company said.

OpenAI said organisations are using the system to summarise medical evidence, draft documentation and adapt patient education materials, while clinicians retain decision-making authority.

All OpenAI for Healthcare products are powered by GPT-5.2 models. The physicians have reviewed more than 600,000 model outputs across 30 focus areas, with feedback used to inform training, safety measures and product updates. The company said ChatGPT for Healthcare also underwent multiple rounds of physician-led testing and evaluation.

OpenAI cited evidence from live deployments, including a study with Penda Health, which found that an AI-powered clinical tool used in routine primary care reduced diagnostic and treatment errors. It also pointed to results from HealthBench, a clinician-designed evaluation that measures model behaviour across realistic medical scenarios, including reasoning, safety and communication quality.

The company said it plans to continue working with healthcare organisations using the new tools to refine the products based on real-world use.