OpenAI opens ChatGPT Health to all US users

OpenAI on July 23, 2026, made ChatGPT Health available to every U.S. user aged 18 and older, pushing a feature that had been in limited testing since January onto all four subscription tiers — Free, Go, Plus, and Pro — as well as the web and iOS.

The nationwide launch arrives at a fraught moment for the San Francisco-based company. Two lawsuits filed in the previous three months allege that ChatGPT dispensed dangerous — in at least one case allegedly life-threatening — medical advice, according to reporting by Gizmodo. One suit, brought by a Florida pastor, claims the chatbot actively discouraged him from seeking professional care. OpenAI has not publicly responded to the specifics of either case, and a company spokesperson did not respond to a request for comment on the pending litigation in time for publication. The timing underscores a tension that physicians, patient-advocacy groups, and regulators have yet to resolve: whether a general-purpose AI assistant can be made safe enough for the messy, high-stakes domain of personal health.

What ChatGPT Health actually does

ChatGPT Health is a dedicated section of the ChatGPT interface — accessible from the sidebar on the web and within the iOS app — that connects the model to a user’s personal health data. According to OpenAI’s official product page, the feature can link to Apple Health (Apple’s iOS health dashboard that aggregates data from an iPhone, Apple Watch, and compatible fitness apps) and to supported medical-records systems. Users who grant access can then ask questions grounded in their own biometrics: sleep patterns, step counts, lab results, medications, and allergies stored in those systems.

Before using health information in any conversation, ChatGPT Health prompts users for explicit permission. The options are one-time access, permanent access, or no access at all — a consent model OpenAI says is backed by “purpose-built encryption and isolation” designed to keep health conversations compartmentalized from the rest of a user’s ChatGPT history, as stated on the company’s product page.

OpenAI said it had been building toward this rollout for more than two years, working with over 260 physicians who have practiced across 60 countries and dozens of specialties. That physician panel, the company says, reviewed model outputs more than 600,000 times across 30 clinical focus areas. The company reported, in its own figures, that more than 300 million people already ask health and wellness questions on ChatGPT each week — roughly 30 percent of the platform’s weekly active users — meaning the behavior was widespread well before a dedicated health product existed.

The initial limited release, announced January 7, 2026, had also connected to MyFitnessPal and other wellness platforms, according to Forbes reporting on the January launch. OpenAI has not specified which third-party integrations carry over to the full rollout, and the company’s July announcement did not address that point.

What it cannot do — and what physicians say

OpenAI’s own language is explicit about the feature’s ceiling. “ChatGPT is not a doctor and should never be used as a substitute for medical care, diagnosis, or treatment,” a company spokesperson told The Register in a written statement accompanying the announcement. The official product page adds that the feature “can still make mistakes,” advising users to confirm anything clinically significant with a qualified professional.

Those caveats have not fully reassured the medical community. Dr. Carolyn Kaufman, a resident physician at Stanford Medicine, told CNN in April 2026 that patient data flowing into AI chatbots carries risks that extend beyond individual health advice. “Data is money,” Kaufman said, noting that she has never uploaded Health Insurance Portability and Accountability Act (HIPAA) — the U.S. federal law governing the privacy of patient health information — protected data to an unapproved chatbot, and warning that doing so “is obviously a risk for the individual patient and for the institution.” OpenAI has stated that ChatGPT Health does not train on users’ health data by default, but it is unclear whether that commitment meets HIPAA’s formal business-associate requirements for covered entities.

The broader concern among clinicians is less about data privacy and more about diagnostic overconfidence. When a patient arrives having already parsed their lab results through an AI, physicians say the conversation shifts in ways that are not always productive. Patients may fixate on AI-generated interpretations that are plausible but wrong, or dismiss a physician’s assessment that contradicts what the chatbot said. The American Medical Association’s 2026 survey of clinician AI usage — cited in the HealthBench Professional research paper published on arXiv — found that clinician use of ChatGPT more than doubled in the past year, which suggests the tool has embedded itself into workflows the medical establishment has barely begun to govern.

The U.S. Food and Drug Administration has not designated ChatGPT Health as a Software as a Medical Device (SaMD) — a regulatory classification that would subject it to clinical validation requirements — and it is unclear whether the agency intends to pursue that classification. The FDA did not respond to a request for comment by publication time.

Market context and what comes next

OpenAI’s move places it more directly in competition with health-focused AI products from Alphabet’s (GOOGL) Google, which has promoted its Med-PaLM 2 model for clinical settings, and Apple (AAPL), whose health-data infrastructure is now a connective tissue for ChatGPT Health’s personalization layer. Neither Google nor Apple commented on the competitive implications of the rollout.

The Dataconomy news service, covering the announcement, noted that the rollout follows a pattern in which health AI products outpace the regulatory frameworks designed to oversee them. OpenAI’s product page acknowledges that ChatGPT Health was built with physician input but stops short of claiming clinical validation under any published benchmark or regulatory standard.

OpenAI’s physician-development panel and its 600,000-plus feedback sessions represent a meaningful investment in safety testing, and the consent-first architecture is more deliberate than what most general-purpose AI products offer. But with two active lawsuits pending and the FDA’s posture on consumer health AI still undefined, the nationwide rollout is as much a legal and regulatory stress test as it is a product launch.

The next milestone to watch is whether the FDA issues formal guidance on consumer-facing AI health tools — agency officials have signaled such guidance is under consideration, though no publication date has been announced. The outcome of the two pending lawsuits against OpenAI will also shape how the company positions disclaimers and consent flows in future updates to ChatGPT Health.

Reporting for this article drew on OpenAI’s official product page, The Register’s coverage of the July 23 announcement, Gizmodo’s reporting on the concurrent litigation, Forbes’s account of the January 2026 limited launch, Dataconomy’s rollout coverage, and CNN’s April 2026 feature on physician reactions to AI chatbots in clinical settings.