AI Is Denying Medicare Claims — and No One's Watching
September 25, 2026The same week Microsoft compares its new AI app to Office, the Trump administration is using AI to deny Medicare claims. The technology is the same category of thing — and the distance between those two applications tells you everything about whose problems AI is actually being built to solve.
The AI Deciding Whether Your Parents Get Medical Care
The Trump administration has deployed AI systems to help process and deny Medicare and Medicaid claims for seniors, and early results are, by multiple accounts, a disaster. The core problem is structural: the vendors building these tools get paid more when they deny claims, so the AI is effectively optimizing for rejection. Seniors are losing coverage for nursing home stays, physical therapy, and other care they're legally entitled to, often without meaningful human review. If you have aging parents navigating Medicare, or you work in healthcare administration, benefits consulting, or insurance, it's happening to real patients right now — and the appeals process is so backlogged that by the time a denial gets reviewed, the care window has already closed.
Microsoft Wants Copilot to Be Your Whole Work Life
Microsoft officially launched its redesigned Copilot "super app" today — a single interface that combines AI chat, coding assistance, and agents (software that can take actions on your behalf, like booking a meeting or filing a report, without you clicking through each step). Comparing it to the launch of Office isn't visionary — it's the kind of grandiose framing you reach for when you need enterprise IT departments to treat a productivity add-on as a platform shift. Whether or not the analogy holds, if you're already inside the Microsoft ecosystem — Outlook, Teams, Word — this is going to touch your daily workflow whether you opt in or not. Before IT enables agents company-wide, it's worth getting specific about what they're actually permitted to do: which systems they can access, what actions they can take without confirmation, and who's accountable when they get it wrong.
https://www.theverge.com/news/1000532/microsoft-copilot-super-app-chat-coding-autopilot
Apple's Home Cameras Got Smarter. Here's What That Actually Means.
Apple updated its home security camera features with AI that's supposed to tell you something useful — "your kid got home from school" — instead of just "motion detected." The Verge tested Apple's version against Amazon Ring and Google Nest to see which one actually understands what it's seeing. The deeper question is whether on-device AI (processing that happens on the camera itself, not sent to a company's server) can match cloud-based accuracy while keeping footage off Amazon's or Google's infrastructure. For anyone advising clients on smart office or retail camera systems, that's not academic: a system that routes footage to a vendor's cloud for analysis is a data-sharing agreement your client probably hasn't read.
The Pentagon Wants an AI Lie Detector. That Should Worry You.
The Department of Defense has requested $30.3 million to build what it's calling Polygraph+, an upgraded lie detection system that uses AI and machine learning to score whether someone is being truthful. The traditional polygraph — which measures heart rate, breathing, and sweat — has been widely discredited by scientists for decades; courts generally won't accept its results. Layering AI on top of a flawed foundation doesn't fix the foundation. The concern for non-defense professionals is the precedent: surveillance and scoring tools developed under defense contracts don't stay in defense. The same contractors selling Polygraph+ to DoD will be selling a repackaged version to corporate HR departments within five years. That's not a prediction — it's the business model.
https://www.technologyreview.com/2026/09/25/1145144/pentagon-ai-lie-detector/
The AI that denies your parents' Medicare claim and the AI that scores whether you're lying to the government are both live and neither has earned the trust being placed in it. The gap isn't between promise and reality. It's between who these systems are designed to protect and who they're being used against.
Stop staring at the blank ChatGPT box
The Plain-English AI Prompt Pack: 100 copy-paste prompts for real work — emails, meetings, marketing, managing people. Works with any chatbot.
Get the Prompt Pack →Get the daily brief
Plain-English AI news for people with real jobs. Free, three minutes a day.