OpenAI's $39 Billion Gamble: Inside the Race to Deploy ChatGPT Across America's Hospitals
Despite securing eight major health systems and mobilizing 260 physicians for quality control, the AI giant faces mounting evidence that its chatbot fails to recognize medical emergencies over half the time.
Illustration: AI-generated
OpenAI is betting its future on healthcare, launching an aggressive expansion that has placed ChatGPT into the workflows of millions of clinicians and the hands of 230 million weekly users worldwide. With three healthcare-focused products released in just six months and partnerships spanning major hospital systems to laboratory giants, the company is positioning itself as the infrastructure layer for medical AI—even as independent research reveals alarming safety gaps.
The stakes couldn't be higher. Healthcare represents 18% of the U.S. economy, and OpenAI—which earned $13 billion in revenue during 2025 while losing approximately $39 billion—confidentially filed for an IPO in June 2025. CEO Sam Altman has personally courted hospital executives, securing enterprise customers including Cedars-Sinai Medical Center and HCA Healthcare in January 2025.
The Three-Pronged Healthcare Strategy
OpenAI's healthcare push operates on three distinct levels. For consumers, ChatGPT Health integrates medical records from Apple Health and wellness apps like MyFitnessPal, offering personalized health coaching currently available only via waitlist. For clinicians, a specialized ChatGPT version targets documentation and administrative tasks—with AdventHealth claiming an 80% reduction in administrative work. At the enterprise level, hospital systems deploy OpenAI tools across their operations, with Memorial Sloan Kettering testing the technology with approximately 1,000 employees.
The company has mobilized over 260 physicians to review medical responses, examining more than 700,000 example interactions to improve accuracy. Healthcare partnerships extend beyond hospitals: Labcorp launched an AI-powered lab results app with OpenAI in May 2025, while healthcare startups Abridge, Ambience, and EliseAI build products on OpenAI's models.
When AI Misses the Emergency
The optimistic deployment timeline collides with troubling clinical evidence. A Nature Medicine study found ChatGPT Health failed to recommend hospital care in over 50% of cases where it was medically necessary. The system failed to identify suicidal patients in some instances while simultaneously over-recommending emergency care for non-urgent situations—a combination that could miss life-threatening conditions and overwhelm emergency departments.
Real-world cases illustrate both potential and peril. Lauren Bannon, a North Carolina patient, reported in January 2025 that ChatGPT suggested Hashimoto's disease when doctors had diagnosed rheumatoid arthritis. Subsequent testing revealed thyroid cancer requiring thyroidectomy and lymph node removal. Dr. Jinsy Andrews, a neurologist, noted that while ChatGPT can prompt earlier care-seeking, it frequently increases patient anxiety with false suggestions.
Regulatory Pressure Mounts
OpenAI has received subpoenas from multiple U.S. states regarding health data handling and safety practices. The company faces a lawsuit from a Canadian woman whose daughter's suicide was allegedly encouraged by ChatGPT interactions. These legal challenges emerge as the company seeks to go public, creating uncertainty about regulatory approval for its healthcare ambitions.
Healthcare AI startups report mixed results with OpenAI's technology. Davis Liang of Abridge noted that OpenAI performs well for clinical note-taking but remains less accurate for detailed diagnoses or billing codes—suggesting the technology may be production-ready for documentation but requires further validation for high-stakes clinical decisions.
The Access Versus Safety Equation
OpenAI's healthcare expansion targets a genuine need: over 33% of rural U.S. adults use emergency rooms for primary care, and 86% of rural counties lack cardiologists. In underserved areas, imperfect AI assistance might seem preferable to no care at all. Yet hospitals operating on thin margins face difficult calculations about competitive pricing pressure, productivity gains, and liability risks from AI-generated errors.
As OpenAI races toward its IPO with healthcare as a cornerstone growth strategy, the fundamental question remains unanswered: Can the company resolve the accuracy and safety concerns quickly enough to justify the scale of deployment already underway in America's hospitals?
Sources
Companies in this article
Reporting & accountability
How we maintain accuracy
Health AI Central separates reported facts, analysis, and public-source data. Readers can review our standards, request a correction, or consult the relevant research methodology before relying on a claim.