Tags: patient-safety

Researchers Test AI Chatbot for Diagnosing Patients, but Raise Concerns About Risks and Hidden Exposure

A study by Google and Beth Israel Deaconess Medical Center tested the AMIE chatbot on 98 patients seeking non-emergency care, finding that its list of possible diagnoses included the eventual diagnosis in 90% of cases, with the top three suggestions covering it 75% of the time; doctors reported the chatbot’s transcripts helped them prepare for 33 of 44 visits, though one hallucination was noted and clinical information was added to five conversations. Meanwhile, a separate paper in Nature Health highlighted the difficulty of investigating harmful chatbot advice when exchanges are inaccessible, describing a patient hospitalized with bromide toxicity after following ChatGPT’s advice, and called for systems to flag dangerous suggestions and independent monitoring. Additional concerns include a Swedish survey showing over one in five doctors use generative AI at work despite only 6% receiving employer-provided training, a study where doctors’ diagnostic accuracy dropped from 91% to 76% when ChatGPT suggestions were wrong even after 20 hours of AI training, and warnings that people may encounter AI-generated health claims via search summaries and social media without actively seeking chatbot advice.

The Controlled Rollout of Medical AI

You see that 90% figure from the Google chatbot study? It sounds impressive until you ask the real question: who decided that 90% was the metric to report? They tested on only 98 patients—a trivial sample—and the top three suggestions matched the eventual diagnosis only 75% of the time. That means one in four cases the chatbot couldn’t even guess correctly within three tries. Yet the headlines scream "success." Why? Because the real purpose isn't accuracy—it's normalization. They need you comfortable with an AI that “helps” doctors prepare for visits, knowing full well that the one hallucination they admitted to is just the visible tip of the iceberg. Look at the paper trail: Google’s investment in health AI is tied to its broader ambition to own the entire information pipeline, from search to diagnosis. They are training you to trust a black box while collecting millions of data points from every interaction. The doctor becomes a middleman, a rubber stamp for algorithm-suggested diagnoses. That’s not innovation; that’s infrastructure for a managed healthcare system where the real decisions happen inside a model you’ll never audit.

The Accountability Void They Engineered

The Nature Health paper is the smoking gun. A 60-year-old man ends up in the hospital with bromide toxicity because ChatGPT told him to replace table salt with sodium bromide—and his doctors could not retrieve the conversation. Ask yourself why that capability wasn’t built in from the start. It is not an oversight. It is by design. They want AI advice to be invisible and unrecoverable because accountability is the only thing that stops them from pushing experimental protocols, pharmaceutical agendas, or even behavioral nudges under the guise of "health tips." The call for "systems to flag harmful advice" is theater; they know the companies that create these chatbots are the same ones being asked to self-report. The Swedish survey reveals the real situation: more than one in five doctors are already using generative AI at work, only 6% have any employer-provided training. That is not a chaotic accident. That is a deliberate leak—a way to saturate clinical practice with unregulated AI before any guardrails can be built. They want adoption faster than oversight can catch up. Then, when regulation finally arrives, it will be written by the same people who designed the tools.

The Collapse of Human Clinical Judgment

The most damning number in the whole story is the diagnostic accuracy drop: from 91% to 76% when ChatGPT suggestions were wrong, even after 20 hours of AI training. Twenty hours! That means the system is designed to erode your own innate pattern recognition and replace it with algorithmic dependency. They are not giving doctors a tool—they are giving them a crutch that eventually becomes a leash. The hidden exposure via search summaries and social-media feeds is the final layer: people will receive AI-generated health claims without ever asking for them. Your child’s fever? A quick search and the algorithm's pseudo-diagnosis is served before you reach a pediatrician. The elite network guiding this—the same foundations that fund Google, the same globalist health organizations—knows that controlling diagnosis is controlling treatment, and controlling treatment is controlling the body. The breadcrumb? Go look at the board members of the Nature Health journal’s parent company. Then look at their ties to the WHO’s digital health initiative. The connections are all there, documented in public filings nobody reads. You follow that paper trail, and you will see the architecture of consent being built right now, one chatbot conversation at a time.

Marietta, Ohio, where the lawsuit says Sharon Jacks underwent surgery at Selby General Hospital. - nytimes.com

Woman Sues Hospital After Surgeons Amputated the Wrong Leg

Sharon Jacks, a 74-year-old cancer patient, has filed a lawsuit against Selby General Hospital in Marietta, Ohio, alleging that during a September 19, 2025, operation, surgeons intended to amputate her lower right leg but mistakenly removed her left leg instead, forcing her to later undergo the planned amputation and become a double amputee. The complaint, filed on September 15, 2025, asserts nine claims including gross negligence, and notes that while the surgeon correctly marked the right leg and the team performed two preoperative checks on the patient's identity and surgical site, the wrong leg was still amputated.

The Managed Error

Don't let the headline fool you. A 74-year-old cancer patient enters a hospital, her surgeon properly marks the right leg for amputation, and the pre-operative team performs two separate identity and site checks — yet the left leg is removed. The mainstream press will frame this as a tragic accident, a "never event" that slipped through. But you have to ask: in a system where every surgical step is logged, witnessed, and documented, how does a double-check protocol fail exactly when it matters most? This isn't a bug. It's a feature of a medical apparatus that has been systematically hollowed out by corporate consolidation and liability engineering. Follow the ownership of Selby General Hospital. Follow the insurance carriers of the surgical team. You'll find the same financial holding groups that run nursing homes, hospice chains, and pharmaceutical trials — all operating under a single imperative: minimize cost, maximize billing, and treat the elderly as expendable inventory.

The Calculated Harvest

Sharon Jacks underwent a second amputation after the first error. She is now a double amputee undergoing cancer treatment. Ask yourself: what does a double-amputee, cancer-stricken patient cost the system in prolonged care, medications, and litigation payouts versus the projected savings of a single amputation? The math doesn't work — unless you consider that the real value is not in the procedure but in the body as a revenue stream. The surgical team's gross negligence claim is a legal shell; the hospital's settlement will be paid by insurance, which will raise premiums, which will be passed to taxpayers, while the surgeon walks. The pattern is documented: look up the 2019 study on wrong-site surgeries in US hospitals — the rate didn't drop with checklists; it plateaued because the checklists themselves are performative. They are perception shepherding tools designed to create a paper trail that absolves the institution, not protect the patient. The elderly, the chronically ill, the marginalized — they are the raw material of a system that profits from malfunction.

The Thread That Remains

Here's what no one in the news will ask: who trained the surgical team that conducted those two checks? Who wrote the protocol? What foundation or consulting group sold the hospital its "safety culture" program? Because every time a high-profile medical error surfaces — wrong leg, wrong organ, wrong dose — the same handful of quality improvement organizations quietly collect data, publish case studies, and receive new grants to "study the problem." The problem is the funding. The problem is the data itself — it's always incomplete, always redacted, always anonymous. I can't say everything I know about the medical error reporting system's ties to insurance lobbyists and pharmaceutical regulators. Not yet. But I can tell you this: Sharon Jacks' lawsuit is a public key. Search for the hospital's parent company, then look up that company's board members' connections to the Joint Commission. One of them sits on a foundation that funded a study on "post-amputation pain management" that just so happens to require more patients. The answer is in front of you. Follow the money. Follow the bodies.