OpenAI Opens Health Feature Nationwide as Liability Questions Mount
The company's health-query tool now reaches all US adult users, even as a lawsuit highlights the gap between AI utility and medical accountability.
A National Rollout Shadowed by Legal Risk
OpenAI has expanded access to ChatGPT Health across all US users over the age of eighteen, a move that arrives one day after a Florida pastor filed suit alleging the platform provided advice that nearly proved fatal. The timing underscores the tension between consumer demand for accessible health information and the legal and clinical boundaries AI systems still struggle to observe.
At DailyTechWire, we've tracked the rise of health-focused AI features across the industry. The pattern is consistent: companies build toward utility, then race to contain liability when real-world consequences emerge. OpenAI's decision to widen availability even as litigation unfolds suggests confidence in its disclaimers and model improvements, but the gap between technological capability and medical responsibility remains wide.
From Pilot to Platform
The feature began as a limited test in January, when OpenAI opened a dedicated health hub that allowed users to connect data from Apple Health, MyFitnessPal, and Function. At launch, the company reported 230 million health-related queries per week. That figure has since climbed to 300 million, a 30 percent increase that reflects both growing usage and the breadth of questions users now direct at conversational models.
Users can now pull in medical records from Epic and Oracle Health, as well as data from One Medical and Function Health. Initially, health queries had to be routed through the dedicated hub. With the nationwide release, users can invoke connected health information in any conversation, allowing the model to offer insights on allergies, nutrition, or medication interactions without switching contexts.
According to OpenAI, 70 percent of health-related queries during the testing phase occurred outside the dedicated hub. That behavioral data informed the decision to let health context flow into general chat, rather than forcing users into a separate workflow.
Model Improvements and the HealthBench Benchmark
OpenAI has pointed to performance gains in its latest model family as justification for broader deployment. The smallest variant in the GPT 5.6 series, referred to internally as Luna, outperforms its predecessor on HealthBench, an open-source evaluation framework the company developed to measure large language model accuracy on health queries.
HealthBench tests models against clinical scenarios, symptom descriptions, and drug interaction questions. While OpenAI has not published granular accuracy metrics, the company emphasized that Luna demonstrates measurably better reasoning on ambiguous or multi-step health questions than GPT 5.5.
The company also stated that it does not use health-related user data to train its models and collaborates with physicians to refine responses. That commitment is central to its privacy posture, but it also limits the feedback loop that might otherwise help the model learn from real-world outcomes.
The Liability Disclaimer and the Florida Lawsuit
Despite performance improvements, OpenAI's terms of service explicitly state that its offerings are "not intended for use in the diagnosis or treatment of any health condition." The company invoked that clause in its response to the Florida lawsuit, which alleges that a user received advice discouraging a visit to a physician, leading to a medical emergency.
OpenAI told media that it is working to make health and medicine answers safer, but the disclaimer remains the company's primary legal shield. The tension is structural: users increasingly treat conversational AI as a first point of consultation, yet the models remain probabilistic systems that cannot guarantee clinical accuracy.
Multiple studies have shown that AI chatbots are unreliable for medical advice, particularly in edge cases involving rare conditions, polypharmacy, or populations underrepresented in training data. Yet demand for accessible health information continues to outpace caution, and companies are responding.
Industry-Wide Movement Toward Health Features
OpenAI is not alone. Anthropic has introduced health-focused capabilities in Claude, and Google has integrated medical knowledge into Gemini, including partnerships with hospital systems and payer networks. The rationale is similar across vendors: people are already asking health questions, and the alternative is unstructured web searches that may surface lower-quality or commercially motivated content.
The difference lies in how companies frame responsibility. Some emphasize decision support for clinicians rather than direct-to-consumer advice. OpenAI's approach, by contrast, is consumer-first, with integration points designed for personal health tracking rather than clinical workflow.
That choice expands the addressable market but also raises the stakes. When a patient uses a tool intended for clinicians, misuse is easier to identify. When a consumer uses a general-purpose assistant that happens to have health features, the line between information and advice blurs.
What Nationwide Access Means for Risk and Regulation
The rollout to free, Go, Plus, and Pro subscribers on web and iOS means millions of users will now have access to health-contextualized responses without additional vetting or onboarding beyond age verification. There is no gating based on health literacy, no requirement to acknowledge risk, and no mechanism to ensure users understand the limitations of probabilistic reasoning in clinical contexts.
From a regulatory standpoint, the feature exists in a gray zone. It is not a medical device under current FDA definitions, nor is it a diagnostic tool. It does not claim to replace physicians. Yet it integrates personal health data, offers interpretations, and influences user behavior in ways that have downstream clinical consequences.
At DailyTechWire, we expect this gray zone to narrow. The Florida case is unlikely to be the last, and a pattern of adverse outcomes could invite scrutiny from the Federal Trade Commission or state health regulators. The question is whether OpenAI and its peers can improve model reliability faster than litigation and regulation can constrain deployment.
A Bet on Verification and Professional Consultation
OpenAI has urged users to verify information and base medical decisions on professional advice. That guidance is sound, but it also shifts responsibility to the user. In practice, many will treat the model's output as authoritative, particularly when it integrates their own health data and offers seemingly personalized recommendations.
The company's emphasis on verification reflects a broader industry hope: that users will treat AI as a research assistant rather than an oracle. But behavior does not always follow intent, especially when the interface is conversational and the model confident.
The expansion of ChatGPT Health is a signal that OpenAI believes the risk is manageable and the demand real. Whether that calculus holds depends on how the next wave of users engages with the feature, how often the model's limitations lead to harm, and how quickly the company can close the gap between what the technology can do and what users expect it to do.
