Microsoft Opens Copilot Health Preview in a Crowded AI Race

Microsoft opened Copilot Health to a US public preview this week, giving Microsoft 365 Personal, Family, and Premium subscribers aged 18 and over a dedicated space to link wearables, lab results, and medical records from more than 50,000 provider organizations, then ask an AI assistant what the numbers add up to. Access is free with an existing subscription, US only, and closed to work accounts.

That headline reads like a single product launch. Look at the calendar and a different picture emerges: Microsoft is the latest of several technology giants to ship a consumer health hub since the start of the year. The chatbot is the part you see. The data it pulls in is the part everyone is fighting for.

What Microsoft Opened to US Subscribers This Week

The preview lives at Copilot.microsoft.com/health and sits apart from the general Copilot assistant. Microsoft first detailed the Copilot Health plan back in March, opened a waitlist, and ran a testing phase with what it describes as thousands of early users before moving to this wider preview after safety evaluation.

Who Can Use It

Eligibility is narrow on purpose. You need a paid Microsoft 365 consumer plan, a US location, and an age of 18 or older. Microsoft has been clear that work and school accounts do not qualify, which keeps the experience squarely in the consumer lane and away from the rules that govern enterprise health software.

What It Connects To

The core idea is consolidation. Instead of leaving your sleep score in one app and your cholesterol panel in a hospital portal, Copilot Health tries to read both in one place and talk through what they mean together. The current preview offers a defined set of building blocks:

• A health profile where you enter background, conditions, and goals so responses reflect your situation rather than a generic case.
• Wearable and record connections, starting with Apple Health and combining with records drawn from more than 50,000 US provider organizations.
• Health insights that ask follow-up questions and suggest what might be going on and what to do next.
• Care navigation that searches local clinicians by specialty, language, gender, insurance, and location.

Microsoft frames this as a deliberate, phased rollout, with more wearable integrations and features promised in the months ahead.

Five Health Hubs in Five Months

Here is the trend the launch coverage tends to skip. Since January, nearly every company with a frontier AI model has put a consumer health product into the market, and most of them share the same blueprint: connect your records and wearables, then answer your questions on top of that combined data.

OpenAI moved first. Amazon followed by widening a health assistant from One Medical members to a broader base. Google laid groundwork with a data-platform partnership last autumn but has not shipped a dedicated Gemini health feature. Anthropic announced healthcare offerings that reach into personal health integrations. Microsoft now joins with the broadest distribution of the group, sitting inside a subscription that already ships on hundreds of millions of PCs.

So the question is not whether an AI company will read your health data. It is which one becomes the default place you keep it.

The Aggregation Layer Is the Prize

The chatbot is replaceable. The data store underneath it is sticky. Once a person links a decade of lab results, a wearable history, and records from multiple hospitals to one assistant, the cost of moving to a rival climbs fast. That switching friction, not the cleverness of any single answer, is what these products are really competing to build.

Most of this connection runs on the same plumbing across vendors. Patient portals expose records through a shared health-data standard called FHIR (Fast Healthcare Interoperability Resources, the format that lets apps read hospital records in a consistent way), which is why so many of these hubs can claim tens of thousands of provider connections. The differentiation moves up the stack to trust, sourcing, and reach.

• 50 million health questions a day already flow across Microsoft’s consumer products, by the company’s own count.
• 50,000-plus US provider organizations are reachable for record import in the preview.
• 230 million people globally ask health and wellness questions on ChatGPT each week, OpenAI has said.
• 3 in 5 US adults reported using AI tools for health purposes in a recent three-month window.

Demand, in other words, is not the constraint. Whoever earns the place where that demand lives wins a position that compounds for years.

Where the Safety Math Gets Uncomfortable

The growth case has a hard counterweight, and it landed in print before Microsoft’s preview. In a study fast-tracked online in Nature Medicine in late February, Mount Sinai researchers ran the first independent safety evaluation of a major consumer health chatbot and found it under-triaged 52% of genuine emergencies, often steering people away from urgent care when they needed it most.

The method was not casual. Investigators built 60 clinician-authored vignettes across 21 clinical domains, then tested each under 16 varied conditions covering factors like race, gender, insurance status, and transportation, for 960 total interactions. They also found the system’s suicide-crisis safeguards fired inconsistently, missing some high-risk scenarios. The Mount Sinai findings on AI medical triage were aimed at a competitor’s tool, not Copilot Health, but the architecture is similar enough that the warning travels.

Microsoft is careful with its own framing. Copilot Health, the company states plainly, is not intended to diagnose, treat, or prevent disease and is not a substitute for professional medical advice. That disclaimer is honest. It is also a reminder that the more a tool knows about your full medical history, the more its confident-sounding suggestions can feel like advice a reader will act on.

The HIPAA Gap Microsoft Designed Around

There is a regulatory fact buried under the privacy reassurances that most readers will miss. Copilot Health is a consumer product, which means it is not a covered entity under HIPAA and carries no business associate agreement. The federal health-privacy law that governs hospitals and insurers does not reach a direct-to-consumer assistant, so the data you hand it is protected mainly by Microsoft’s own commitments and a patchwork of state privacy laws.

That is not unique to Microsoft; the same gap applies to its rivals’ consumer hubs. It matters because concentrating medical records, wearable streams, and AI chat logs in one account creates a single high-value target, and the usual breach-notification and security mandates that come with HIPAA simply do not attach here.

Microsoft’s answer is to build protections by contract and design rather than by statute. Health conversations are walled off from the rest of Copilot, are not used to train its AI, and are encrypted in transit and at rest. Users can delete or disconnect any data source at any time. Those are real safeguards. They are also promises a company makes, not obligations a regulator enforces, and the distinction is worth understanding before you link your records.

How Microsoft Is Trying to Earn Trust

The pitch leans heavily on credibility borrowed from medicine. Copilot Health draws its guidance from thousands of trusted health organizations and from a partnership with Harvard Health, with sourcing principles independently published by the National Academy of Medicine. Microsoft says it built the product with its internal clinical team and an external panel of more than 250 physicians from over 24 countries.

It has also pursued formal certification. Copilot Health holds ISO/IEC 42001 certification (an international standard for AI management systems), meaning an outside auditor reviewed how Microsoft builds and governs the underlying AI. And the company brought in advocacy groups to vet the experience, including members of the National Health Council, which offered a measured endorsement.

By engaging patients, caregivers, and advocates into the process from the start, the program is shaping a solution grounded in real-world needs and helping build lasting confidence in AI-enabled care.

That statement came from the National Health Council, a coalition representing patient advocacy groups, reflecting on its work shaping the preview. The endorsement is genuine, and so is the gap it cannot close: an organization can advise on design without being able to vouch for every answer the model gives a worried user at midnight.

Frequently Asked Questions

Who Can Use Copilot Health Right Now?

Only US residents aged 18 and over who hold a Microsoft 365 Personal, Family, or Premium subscription can access the preview. Work and school accounts are not eligible, and availability outside the US has not been announced.

How Much Does Copilot Health Cost?

There is no separate fee. Copilot Health is included with an eligible Microsoft 365 consumer subscription. Microsoft notes that features and usage limits may change during the preview period.

Is Copilot Health HIPAA Compliant?

No. As a direct-to-consumer product it is not a HIPAA covered entity and carries no business associate agreement. Microsoft protects the data through its own encryption and privacy commitments and applicable state laws rather than through HIPAA obligations.

What Devices and Records Can Copilot Health Connect To?

The preview starts with Apple Health for wearable and wellness data, with more integrations promised over time. It can also import medical records from more than 50,000 US provider organizations through patient-portal connections.

Can Copilot Health Diagnose a Medical Condition?

No. Microsoft states the tool is not intended to diagnose, treat, or prevent disease and is not a substitute for professional medical advice. It is built to surface insights and help you find care, not to replace a clinician.

Is My Health Data Used to Train Microsoft’s AI?

No. Microsoft says Copilot Health conversations are kept separate from the rest of Copilot and are not used to train its AI models, and users can delete or disconnect their health data sources at any time.

If Microsoft’s trust scaffolding holds up under independent testing and the kind of scrutiny rivals are already facing, the subscription reach could make Copilot Health the default health hub for tens of millions of households. If the accuracy and privacy questions catch up with it first, all that distribution just spreads the risk faster.

Disclaimer: This article is for informational purposes only and does not constitute medical advice. AI health tools, including Copilot Health, are not a substitute for professional diagnosis or treatment; consult a qualified healthcare provider for medical decisions. Product features, eligibility, and figures are accurate as of publication and may change during the preview period.