你所有的睡眠数据、心率记录、步数、卡路里消耗、甚至血液检查结果——都可以直接喂给ChatGPT了。
7月23日，OpenAI正式宣布Health in ChatGPT完成全量上线，覆盖Free、Go、Plus、Pro全部四个套餐，Web端和iOS端同步生效。所有美国18岁以上用户，授权后可以把Apple Health、MyFitnessPal、Function等健康应用数据，以及部分医疗机构的电子健康记录（EHR）接入ChatGPT，让AI基于你的真实健康数据给出个性化建议。
这是OpenAI从"通用聊天机器人"走向"个人健康助理"的关键一步。但上线同一天，一则诉讼新闻给这场发布蒙上了阴影：一位美国牧师起诉OpenAI，称ChatGPT给出的健康建议险些让他丧命。
它能做什么？不能做什么？
根据OpenAI官方介绍，接入健康数据后的ChatGPT可以做这些事：帮你分析睡眠趋势和心率变异性，结合运动和饮食数据给出改善建议；解读你的血液检查报告，用通俗语言解释各项指标意味着什么；基于你的用药历史和健康目标，提醒你注意药物相互作用；甚至在连接医疗记录后，帮你整理就诊历史、为下一次看医生准备问题清单。
OpenAI反复强调：ChatGPT Health不是医疗设备，不能诊断疾病，不能替代医生。界面上会有醒目的免责声明，建议用户在做出医疗决定前咨询专业医生。但问题是——当一个AI工具能看到你所有的健康数据、能给你具体的健康建议时，普通用户真的能分清"这只是健康建议"和"这是医疗诊断"的边界吗？
这正是那起诉讼的核心争议点。据爱范儿报道，一位牧师声称他按照ChatGPT给出的建议调整用药后出现严重健康问题，险些致命，因此起诉OpenAI。OpenAI尚未对此诉讼做出正式回应，但可以预见的是：这不会是最后一起围绕AI健康建议的诉讼。
AI医疗的商业化与责任困境
Health in ChatGPT的上线，标志着AI医疗从"概念验证"进入"大规模商用"阶段。美国医疗市场规模超过4万亿美元，数字健康是其中增长最快的细分领域之一。OpenAI选择健康作为第一个深度垂直领域，商业逻辑很清晰：健康是每个人的刚需，用户粘性高，付费意愿强，数据价值大。
但健康也是监管最严、责任最重的领域。如果你让ChatGPT帮你写邮件写错了，大不了重写；如果你让ChatGPT给你健康建议出错了，可能出人命。这就是为什么OpenAI虽然功能上线了，但在措辞上异常谨慎——反复强调"不是医疗建议""不能替代医生"，本质上是在法律层面给自己留退路。
"当AI能看到你从心跳到血液检查的全部健康数据，它给你的建议就不再是'仅供参考'——用户会本能地信任它。"—— 一位数字健康领域投资人
从产品角度看，ChatGPT Health的想象空间很大。如果未来能接入更多数据源——比如连续血糖监测仪（CGM）数据、智能手表的心电图、甚至基因数据——ChatGPT可以成为一个真正的"全天候健康管家"。Google和Apple在健康领域布局多年，但没有一家拥有OpenAI这样的自然语言交互能力。你不需要打开五个健康APP看图表，直接问ChatGPT"我最近睡眠为什么变差了"，它就能给你综合分析。
但这条路不会好走。HIPAA合规、医疗责任界定、数据隐私保护、误诊风险——每一道坎都是硬仗。更关键的是用户信任：一次严重的误诊事故，就可能毁掉整个产品的口碑。
OpenAI选择在这个时间点全量上线Health功能，既是商业化的需要，也是在抢市场窗口期。Google Gemini、Anthropic Claude都在虎视眈眈健康赛道，先下手为强。但AI医疗的战争，最终拼的不是谁的模型更聪明，而是谁能在安全和责任的边界上走得更稳。
明天见。
All your sleep data, heart rate records, steps, calories burned — even blood test results — can now be fed directly to ChatGPT.
On July 23, OpenAI officially announced that Health in ChatGPT had launched to all users, across Free, Go, Plus, and Pro tiers, on both web and iOS. All U.S. users 18 and up can, after granting authorization, connect health data from Apple Health, MyFitnessPal, Function, and electronic health records (EHR) from select medical providers — letting ChatGPT deliver personalized advice based on your real health data.
It's a critical step in OpenAI's evolution from "general chatbot" to "personal health assistant." But on the very day of the launch, a lawsuit cast a shadow over the announcement: an American pastor is suing OpenAI, claiming that ChatGPT's health advice nearly killed him.
What It Can Do — and What It Can't
According to OpenAI's official materials, ChatGPT Health can do the following once connected to health data: analyze sleep trends and heart rate variability, and suggest improvements based on exercise and nutrition data; interpret blood test reports and explain what each metric means in plain language; flag potential drug interactions based on your medication history and health goals; even, after connecting medical records, help you organize visit history and prepare question lists for your next doctor's appointment.
OpenAI is emphatic: ChatGPT Health is not a medical device, cannot diagnose disease, and is not a substitute for a doctor. Prominent disclaimers appear in the interface, advising users to consult a healthcare professional before making medical decisions. But here's the problem: when an AI tool can see all your health data and give you specific health recommendations, will ordinary users really be able to draw the line between "that's just wellness advice" and "that's a medical diagnosis"?
That tension is precisely at the heart of the lawsuit. As reported by ifanr, a pastor claims he adjusted his medication based on ChatGPT's advice, suffered severe health consequences that nearly killed him, and is now suing OpenAI. OpenAI has not formally responded to the suit, but one thing is predictable: this won't be the last lawsuit over AI health advice.
The Commercialization and Liability Dilemma of AI Healthcare
The launch of Health in ChatGPT marks AI healthcare's move from "proof of concept" to "mass commercialization." The U.S. healthcare market is worth over $4 trillion, with digital health as one of its fastest-growing segments. OpenAI's choice of health as its first deep vertical makes clear commercial sense: health is a universal necessity with high stickiness, strong willingness to pay, and massive data value.
But health is also the most heavily regulated, highest-liability domain. If ChatGPT writes a bad email, you rewrite it. If ChatGPT gives you bad health advice, you could die. That's why OpenAI, despite launching the feature, is being extraordinarily careful with its language — emphasizing repeatedly that it's "not medical advice" and "not a substitute for a doctor," essentially building legal cover for itself.
"When an AI can see everything from your heartbeat to your blood work, its advice stops feeling like 'for reference only' — users will instinctively trust it."— A digital health investor
From a product perspective, ChatGPT Health has enormous potential. If more data sources come online in the future — continuous glucose monitor (CGM) data, smartwatch ECGs, even genomic data — ChatGPT could become a genuine "24/7 health steward." Google and Apple have been laying groundwork in health for years, but neither has OpenAI's natural-language interaction capability. Instead of opening five health apps to look at charts, you can just ask ChatGPT, "Why has my sleep gotten worse lately?" and get a synthesized answer.
But the road ahead is bumpy. HIPAA compliance, medical liability frameworks, data privacy, misdiagnosis risk — each hurdle is a major battle. Most critically is user trust: one serious misdiagnosis incident could destroy the product's reputation.
OpenAI's choice to roll out Health broadly at this juncture is both a commercial necessity and a land-grab. Google Gemini and Anthropic Claude are both eyeing the health space, and first-mover advantage matters. But the AI healthcare war won't ultimately be won by whoever has the smartest model — it'll be won by whoever navigates the line between safety and liability most steadily.
See you tomorrow.
当AI能看到你从心跳到血液检查的全部健康数据，它给你的建议就不再是'仅供参考'。
—— 一位数字健康领域投资人
When an AI can see everything from your heartbeat to your blood work, its advice stops feeling like 'for reference only.'
— A digital health investor
ChatGPT Health · OpenAI · Apple Health · 电子医疗记录 · AI医疗 · 健康建议 · 法律诉讼 · 医疗责任 · HIPAA · 个性化健康
ChatGPT Health · OpenAI · Apple Health · EHR · AI healthcare · health advice · lawsuit · medical liability · HIPAA · personalized health
