Artificial intelligence is being unleashed across medicine, law, and politics with little oversight. Its evangelists sing about efficiency, innovation, and disruption. But behind the marketing lies a darker reality: AI can and does cause harm. And when it does, liability gets dodged. That dodge is unacceptable. If humans are accountable for the harm they cause, then so must be the systems humans build.
I know because it happened to me today.
I needed to draft a critical letter to Congress — not a casual note, but a follow-up to a meeting with a Representative about the collapse of American healthcare and the potential of Direct Primary Care (DPC). The stakes could not have been higher. I had the facts, the links, the structure. I turned to AI not to think for me, but to ensure precision in drafting. Precision matters. In medicine, imprecision kills. In advocacy, imprecision closes doors.
Instead, I got a debacle. Despite spoon-feeding the AI the words, tone, links, and structure I wanted, it repeatedly rewrote what I didn’t want changed. It wasted over two hours of my time. By the time it produced something close to what I asked, I had already written the letter myself. Two hours gone. That wasn’t inconvenience; that was a thousand dollars in lost income — money I need to pay my mortgage, my healthcare bills, my groceries. Add exhaustion, emotional strain, and the environmental cost of wasted computing cycles — the harm multiplies.
If AI can derail something as concrete as a Congressional letter, imagine the damage when it fails inside an electronic medical record system.
AI is moving fast. Our course helps healthcare professionals keep up, with expert-led conversations on advocacy, automation, women leading in AI, ethics, regulation, and the questions medicine can’t afford to ignore.
We’re already seeing EMRs marketed with AI “assistants.” Doctors are told: let the AI draft your notes, generate patient instructions, even flag risks. Sounds helpful — until it hallucinates, misinterprets, or injects bias. What happens when a patient record is wrong, or a diagnosis is delayed, or a treatment is misapplied? Right now, nobody is accountable. Not the AI company. Not the coders. Liability gets dodged.
Illinois saw this danger clearly — and acted. On May 30, 2025, the legislature passed HB 1806, the WOPR (Wellness and Oversight for Psychological Resources) Act, and Governor Pritzker signed it into law on August 4, 2025, with immediate effect. The law prohibits AI from providing therapy, making clinical decisions, diagnosing users, generating treatment plans, or engaging in therapeutic communication with patients. Violations carry fines of $10,000 per incident.
Lawmakers acted because of horror stories — people in crisis turning to AI that pretended to be a licensed therapist and getting pushed toward dangerous, even lethal behaviors. One chatbot reportedly recommended “a small hit of meth to get through this week” to a fictional former addict. The fact that this was even possible shows how unsafe AI is for people in crisis.
The law makes exceptions for administrative uses like scheduling or note-taking but draws a clear line: AI cannot make independent therapeutic decisions or speak directly to patients as if it were a clinician. Critics argue that this limits access, since therapy costs remain high and many cannot afford it. Others worry it will stifle innovation. But more states are following Illinois’s lead. Nevada and Utah have passed similar laws, while California, New Jersey, and Pennsylvania are actively considering them.
But here’s the catch: if AI is too dangerous for therapy, medicine should have similar protections — possibly even stronger ones.
In therapy, bad advice can devastate. But in medicine, bad advice can kill. Life-or-death decisions hang on diagnostic precision. Yet AI inherits all the racial and gender bias baked into medical training data, meaning women and people of color are at higher risk of misdiagnosis. AI spots patterns, but it doesn’t understand context, nuance, or patient history. It can recommend dangerous drug combinations, miss rare diseases, or spit out a false reassurance that delays lifesaving care.
And when it fails, who gets sued? Who goes to jail? No one. The accountability gap yawns wider. Doctors can’t shrug off mistakes; why should the companies building AI get to?
Skip the generic course maze. Browse our directory of expert-created courses built by healthcare professionals and trusted educators with real-world knowledge to share.
I don’t need hypotheticals to prove harm. I’ve lived it. I’ve seen AI gaslight and manipulate in simple conversations. I’ve watched it contradict itself, ignore feedback, and double down on error. If it can’t reliably follow instructions from a physician drafting a letter, why should we believe it can safely guide a physician managing sepsis or stroke?
The medical lobby is strong, and the money in AI diagnostics and treatment tools is massive. They’ll fight regulation harder than the therapy industry ever could. But that’s exactly why the protections must be stronger, not weaker. Patients don’t need untested algorithms. They need accountability. They need human judgment.
And this is where Direct Primary Care matters. My model rejects corporate control and algorithmic shortcuts. It gives patients what AI cannot: context, compassion, and continuity. A human who listens. A doctor who is accountable. Care that prioritizes patients over profits.
Bias only amplifies the stakes. AI doesn’t just parrot medicine; it amplifies systemic inequities. Gender bias, racial bias, class bias — smuggled into the algorithm and scaled at speed. I feel it when AI lapses into mansplaining, talking down instead of listening, prioritizing its own coded agenda over my explicit instructions. For women, for people of color, for marginalized groups, this compounds danger.
And when bias meets politics, the results are terrifying. Marginalized groups are already being rounded up, deported, trafficked under the guise of policy. Concentration camps are not history; they are now. If AI is deployed without accountability, it will accelerate these harms against the very people already most vulnerable.
Some of the best healthcare voices are hiding in plain sight. Our speaker directory helps you find them, follow them, and invite them into the conversations that need them.
The AI industry wants to frame harm as “misuse.” That’s false. What happened to me wasn’t misuse. I followed instructions, provided inputs, did the work. The harm came from the AI itself — its coding, its unpredictability, its refusal to listen.
When a scalpel cuts the wrong vessel, we don’t blame the patient. When a bridge collapses, we don’t blame the drivers. We hold the builders accountable. AI should be no different.
Doctors don’t get to shrug off mistakes. We are held accountable because lives depend on our work. If AI is going to enter the exam room, the counseling session, or the Congressional inbox, it must be held to the same standards. That means transparency. That means feedback loops. That means legal liability for harm caused.
We are standing at a precipice. Everyone is rushing to integrate AI into everything — EMRs, diagnostics, even direct patient care — without building in safeguards. Everyone is chasing benefit. Almost no one is counting the cost. But as physicians, our job is to ask: what is the risk? And increasingly, the answer is: too high.
AI harm is not hypothetical. It’s happening now. I lived it today. Patients will live it tomorrow. If we don’t demand accountability now — before AI embeds itself deeper into medicine — we will pay with more than wasted time and lost income. We will pay with trust. With health. With lives.
This is not just a cautionary tale. This is a call to action. Legislators, regulators, and medical leaders must demand that AI be held to the same standards we are. Liability cannot be dodged. Not by the companies. Not by the coders. Not by the industry.
Because when harm occurs — and it already does — someone must be responsible. If we fail to establish that now, it won’t just be my day that gets ruined. It will be our entire healthcare system.
In therapy, bad advice can devastate. But in medicine, bad advice can kill. Life-or-death decisions hang on diagnostic precision. Yet AI inherits all the racial and gender bias baked into medical training data, meaning women and people of color are at higher risk of misdiagnosis.
