AI has spent the last few years drafting emails, summarizing meetings, and writing marketing copy. Now it is moving somewhere with actual stakes attached. OpenAI just rolled out Health in ChatGPT, letting U.S. users connect Apple Health and medical records directly into their conversations. Google, Amazon, and Apple are all building toward some version of the same thing. And on the other end of the system, health insurers have quietly been using AI to help decide what care gets covered, which is exactly why Congress is now getting involved.

Here's how it breaks down for everyone touched by it, patient, platform, insurer, regulator, and yes, even the dealership down the street.

You now have a chatbot that can see your lab work, your Apple Watch data, and your last three visit summaries at once. That is the entire pitch behind Health in ChatGPT, OpenAI's new rollout to U.S. users.

The old version made you go to a separate health tab to get any benefit from connecting your records. OpenAI killed that. Turns out more than 70% of health conversations were already happening outside the dedicated space, mixed into questions about dinner plans or weekend hikes. So now the context just follows you. Mention a knee injury while planning a hike, and ChatGPT already knows about the injury.

Nothing requires a login to Apple Health or a hospital portal. You can still ask a general question with zero data connected. But if you do connect something, every use of it triggers a permission prompt by default (unless you flip that off yourself).

OpenAI is betting its advantage is the model, not just the data pipe. GPT-5.5 Instant, free to everyone, now performs on health questions at a level OpenAI compares to their top "thinking" model from a year ago. GPT-5.6 Sol, the paid version, scores higher still on tasks that require reasoning across a messy medical history, explaining a confusing visit note in plain English, or knowing when to say "call your doctor" instead of guessing.

Hundreds of physicians helped build the scoring rubric behind those claims: accuracy, clarity, completeness, and whether the model escalates properly when something looks serious.

OpenAI is not alone in wanting this real estate. Google has been building AI into Fitbit and Health Connect. Amazon owns One Medical and is threading AI into pharmacy and virtual care behind the scenes. Apple stays closer to the hardware, but every company connecting to Apple Health is really building on top of Apple's plumbing without Apple having to build a chatbot at all.

None of this data trains OpenAI's models or targets ads, according to the company, regardless of your other settings. Disconnect a source and it is gone from their systems within 30 days.

This is where the story gets less comfortable. Health plans have used AI for years to sort and process claims, and increasingly to help decide what counts as "medically necessary." That phrases is doing a lot of heavy lifting.

It is also the word behind a growing number of denials for mental health coverage, cancer treatment, and hospital stays, according to lawmakers who have been hearing from constituents about exactly this.

An insurer using AI to flag a claim for review is a very different thing than an insurer letting AI issue the actual denial. Right now, in a lot of cases, that line is blurry, and patients often have no idea an algorithm touched their case at all.

That blurriness is exactly what a new bipartisan bill, the Doctors Not AI Act, is trying to fix. The bill does not ban AI from insurance operations. It requires that any denial based on medical necessity get a real review from a licensed professional, not just an algorithm's output, and it gives patients the right to know when AI played a role in their denial.

Think of it as regulators drawing the same line the rest of this industry is drawing on its own: AI can help sort, summarize, and speed things up. It does not get to be the final word on anything that takes away care. Expect more state and federal proposals shaped the same way in the next year, not fewer.

None of this is happening in an industry adjacent to yours. It is happening in F&I.

Extended warranties, GAP coverage, and third-party health and wellness add-ons increasingly run on AI-assisted underwriting and claims processing, the same category of tool regulators are now circling on the medical side. Call your F&I providers and ask directly whether AI touches their claims decisions, and what the appeal process looks like when a customer gets denied. If you don't know the answer today, you're not alone, but you should be finding out.

Check your own employee benefits too. If your dealership group offers health coverage through a major carrier, that carrier may already run claims through an AI system. Ask your broker whether denials get a genuine clinical review, and whether that reviewer looks at the individual case or just confirms what the algorithm already decided.

The upside here is real. So is the tradeoff. AI can turn a decade of scattered data into a conversation worth having with your doctor, and it can also become a permanent record of your health history sitting on a server somewhere, subject to whatever happens to the company that holds it next.

If you're rolling out any AI tool that touches customer data, whether that's a service chatbot or a CRM reading purchase history, you're already being asked the same question everyone in this story is answering right now: does the customer know what's happening to their data, and can they actually get it back?