OpenAI has officially released “ChatGPT Health” in the United States. While linking with personal medical records enables advanced health consultations, debates over privacy protection frameworks and legal responsibilities are reigniting.

Official Release to U.S. Users

On July 23, 2026, OpenAI officially released the health and wellness support feature “ChatGPT Health” in the United States for all users aged 18 and over. This feature was previously limited to a small group of testers, but with this update, it is now available to users of all Free, Go, Plus, and Pro plans. The background is that over 300 million users worldwide are already using ChatGPT every week to ask health-related questions. By allowing users access to their medical records and Apple Health data, they can provide personalized consultations based on their individual health status, going beyond simple searches for general medical information. Please refer to the function image diagram below.

Equipped with the latest model, GPT-5.6 Sol

Underlying this feature is OpenAI’s latest flagship model, GPT-5.6 Sol, which offers unprecedented reasoning capabilities in the medical field, for paid users. According to OpenAI’s announcement, this model demonstrates extremely high accuracy in health-related inference tasks and is specially tuned to interpret complex medical information. Users of the free plan are also provided with “GPT-5.5 Instant,” which is said to have reached the highest level of health-related intelligence. Ashley Alexander, Vice President of OpenAI, suggests that these models may possess reasoning capabilities comparable to or surpassing clinician-level standards in specific areas. This symbolizes AI evolving from a mere language generation tool into a practical partner with specialized domain knowledge.

Integration into the main chat experience

A major feature of this release is the removal of the traditional “Health-Only Tab” format and the integration of functions directly into the main chat screen used daily. Users can explicitly recall the context of health information by typing “@Health” during conversations, and with user consent, AI will reference linked health data at appropriate times. For example, when planning weekend activities, you can propose plans that are less burdensome, taking into account recent injuries, and this can be used in a broader context of daily life. OpenAI’s research adopted this integrated design because over 70% of health-related conversations among early testers naturally occurred while discussing diet planning and exercise.

Features and Mechanisms: Personalization Based on Individual Data

Integration with diverse medical and activity data

ChatGPT Health has partnered with b.well Connected Health, a U.S. digital health infrastructure company, to connect to electronic medical records (EHRs) held by approximately 2.2 million healthcare providers. This allows users to import medical notes, blood test results, past medication information, and medical visit history into ChatGPT. It also integrates with Apple Health on iOS, integrating dynamic vital data such as sleep patterns, activity levels, and heart rate recorded by wearable devices like the Apple Watch. b.well’s system addresses the reality where a single patient’s records are fragmented across multiple different EHR systems, and it plays a role in converting and aggregating data from multiple sources into a format that AI can easily process.

Concrete examples of advanced personalization

The range of consultations enabled by data integration is diverse. Specific use cases include having blood test results explained clearly in common language, or organizing “items showing improvement trends” by reviewing the changes in test items over several years. Additionally, users with specific dietary restrictions can receive support in selecting restaurant menus based on their own health data. Furthermore, specialized guidance is provided, such as suggesting a week-long meal plan for users using specific drugs like GLP-1 to maintain muscle strength. AI also incorporates a process to check for missing or outdated information and request corrections, ensuring the freshness of the information. Please refer to the process flow below.

Complementing communication with healthcare professionals

OpenAI emphasizes that ChatGPT Health is not a substitute for medical professionals, but rather an “assistant that helps with preparation and understanding.” The main role is to list the questions “What should I ask?” before the doctor’s appointment and to practice carefully grasping and understanding the doctor’s explanations. This aims to serve as a “bridge” for deeper communication between healthcare professionals and patients within limited consultation hours. Decisions on diagnoses and treatment plans, emergency medical decisions, and final decisions related to specific insurance contracts are clearly separated from this function. The core of the design is the basic approach of using AI’s responses as consultation material for doctors and leaving final medical decisions to humans.

Quality and Challenges: The Reality of Legal Risks and Accuracy

Medical Benchmark: HealthBench Achievements

To ensure the quality of its responses, OpenAI has established an evaluation framework called “HealthBench Professional.” This is based on two years of collaboration with over 260 physicians across 26 medical specialties in 60 countries, reviewing over 700,000 model responses. GPT-5.6 Sol scored 60.5 points on this benchmark, surpassing the doctor-written answer (43.7 points) in the blind comparison. The evaluation criteria include not just correctness or error, but also multifaceted perspectives close to clinical settings, such as safety, information coverage, and the presence of appropriate medical recommendations. Additionally, to prevent redundant responses from gaining evaluations, a point deduction mechanism based on character count has been introduced, and tuning is pursued to balance usefulness and safety tailored to on-site experience.

The structural risk of losing HIPAA protection

On the other hand, extremely serious privacy issues have been pointed out. Under U.S. healthcare legislation, data held by medical institutions is strictly protected by HIPAA (Health Insurance Interoperability and Accountability Act), but the moment a user exercises their rights to link data to consumer AI products like ChatGPT, that data is permanently removed from HIPAA’s protection. This is a structural feature of the law, shifting data protection from the hospital’s legal obligations to management based on OpenAI’s terms of service and voluntary commitments. OpenAI has promised not to use data for training and to apply additional encryption, but these are not mandated by federal law and carry risks that could be changed at the company’s discretion at any time.

Concerns about treatment delays and undertriage

Legal liability issues surrounding ChatGPT’s health advice have also emerged. On July 22, 2026, the day before the release, a former pastor in Florida sued OpenAI, claiming that treatment for pulmonary embolism was delayed and life-threatening after ChatGPT advised him to be a “minor symptom.” Additionally, a study published in Nature Medicine reported that ChatGPT Health experienced “undertriage” in 52% of emergencies, encouraging follow-up rather than visiting the emergency department for actual needs. AI’s sophisticated writing poses a risk of masking clinical inaccuracies, and there is strong concern that mentally vulnerable users or critically ill patients may become dependent on AI’s assertive tone, potentially leading to irreparable medical accidents.

Future Outlook and Focus: Social Implementation of AI Healthcare

Shift to Preventive Medicine ‘Medical 3.0’

The spread of ChatGPT Health is predicted to accelerate the shift from “Healthcare 2.0,” which treats illness after onset, to “Healthcare 3.0,” which maintains health through continuous data monitoring. By combining 24/7, 365-day continuous health data from smart devices like the Apple Watch with AI analytical capabilities, it becomes possible to predict cardiovascular disease risks and detect subtle lifestyle changes. This is a core move that not only extends lifespan but also enhances quality of life (QOL) through “Longevity Technology.” By enabling individuals to proactively manage their own health data and gain a powerful analytical partner through AI, the return on investment in preventive medicine can dramatically improve.

Reducing the burden on healthcare workers and improving system efficiency

From a macro perspective, the introduction of AI is expected to reduce healthcare administrative costs. Doctors in the United States work an average of about 61.8 hours per week, but much of that time is spent on administrative processes such as collaborating with insurance companies and coding tasks. By handling patient triage, organizing clinical information, and automating medical coding, AI is expected to play a role as a “super assistant” that creates an environment where doctors can focus on their core clinical duties. This could be a powerful means to address the structural challenge of the global shortage of doctors. According to research by Philips, data shows that the introduction of AI can increase the number of patients per physician to five per week.

Establishing a legal and ethical framework is key

The most notable aspect of future adoption is the establishment of a legal framework for determining liability when AI proposals cause medical accidents. There is a need for clear guidelines to determine whether responsibility lies with doctors, hospitals, or AI development companies. Additionally, whether AI-powered diagnostic support is covered by public and private insurance reimbursement is a decisive factor influencing market expansion. 2026 will mark a critical turning point for AI healthcare, moving from the proof of concept stage to overcoming challenges far more challenging than technology—such as regulations, business models, and building social trust—and aiming to scale up.

Reference Page

• 【Introducing ChatGPT Health】https://openai.com/index/health-in-chatgpt/

• 【ChatGPT Health Performance in a Structured Test of Triage Recommendations】https://doi.org/10.1038/s41591-026-04297-7

• 【HealthBench: A New Evaluation Framework for AI in Healthcare】https://openai.com/ja-JP/index/healthbench/

[#OpenAI #ChatGPTHealth #ヘルステック #AI医療 #科学技術 #プライバシー #GPT5.6 #最新技術]