June 13, 2026 at 12:00 am
Comments posted to this topic are about the item Liability for AI Errors
June 13, 2026 at 6:14 am
Hot button here. AI has already invaded my medical files , and given me cancer that I have never had. So this spring, I went in to surgery, with the surgery team having medical notes on file telling them that I have serious medical issues that don't exist. Because AI misinterpreted the language that was spoken in a room during a pre-op exam. My doctor clearly said the same word twice, and I just as clearly heard it the same as he said. AI summarized our conversation and inserted the wrong term in my medical record twice. The busy surgeon, if/when he reviewed the record, did not catch the mistake.
So you could say the surgeon is at fault, for not catching the mistake. But then, he isn't the one who made the mistake in the first place. Fortunately, my surgery was successful, and not impacted by this particular mistake. I don't know who should be liable here, but surely, if AI was not used, this particular mistake would never have occurred. I suspect when it comes to litigation, both should be held responsible in a case like this, because there were two different errors here.
Thirty years of experience with databases and code tells me that approximations, even based on millions of examples, are still only approximations. I don't want my life depending on approximations because it saved the doc 10 minutes of time, and I don't think AI should be let off with the excuse that humans make mistakes too.
June 14, 2026 at 3:17 pm
WOW Adrienne, what a story you relayed that happened to you!! And it makes me think, the last time I saw my doctor he did the same thing, in that he asked me if I would allow an AI to listen to our conversation, to take notes. I said yes, but your story shows me that I must be much more cautious!! What happened to you doesn't give me confidence that AI is better than humans at assisting with diagnosing. There is another problem, though. During that meeting the doctor told me that the medical company he works for is experiencing layoffs as well. Last week they had a cut in staff. So, I'm sure the temptation to use AI has increased just because they're cutting staff. This article, your experience, and staff cuts at medical providers suggest to me that misdiagnosis of medical conditions is going to go up. People are playing with fire without even seeing the flames.
Kindest Regards, Rod Connect with me on LinkedIn.
June 14, 2026 at 10:07 pm
Indeed AI creators don't know how it works. They try to mimic structures already observed in the brain, and when a given threshold in layers and complexity of reached, then it starts to work.
Then it comes the training phase, and here it's what liability could happen. But as long as there was no bad intention or bias in the training, I understand the doctors should interpret the results before taking decisions.
June 15, 2026 at 3:29 pm
I believe that the issue here is data vs. information, then actions taken upon the information.
Medical lab equipment: data.
AI/Doctor interpreting data: information.
Surgeon acting upon information.
I'm with you, both should be accountable but separately. A surgeon should be accountable for his actions, AI accountable by the surgeon for turning good data into bad information.
I'm a software engineer. Say I misinterpret data and design controller software based on that data, and the controller causes an airliner to crash. Who's at fault? Me. Who's accountable? Yeah, I'd probably get fired, but I wouldn't be the one going to court...the maker of the airliner would.
Software engineers build an AI. The AI makes a mistake and tells an airliner to turn left, plowing into a mountaintop. Who responsible? The airline. Who caused the accident? The airline's pilots. Who goes to court? The AI makers and the airline, both.
What's the difference? The airline gets sued by the survivors, their families, and the government agency regulating airlines. The makers of the AI get sued by the airline. 1 incident, 2 causes, 2 court cases. Yes, I believe AI companies should be held equally accountable for their mistakes.
Another example? A car passenger is killed in an accident because their seatbelt failed when the driver fell asleep at the wheel and plowed into an oak tree. Who's at fault? Both the driver and the car manufacturer, and maybe the car company then sues the company that made the faulty seatbelts. Seatbelts are the data, here. But you see? The victim's survivors don't sue the seatbelt maker, they sue the car maker who then sues the seatbelt maker.
The point is, everyone involved in an incident is responsible for their portion of the cause. AI is obviously part of the cause of an incident where someone triggering the incident used the AI. I'm looking at AI as a service because that's what it is, although in cases where AI is also the program (i.e. the one performing the action) like Claude Code, Microsoft Agent Framework, etc.
Coder since 1977.
Been there.
Done that.
A lot.
Still doing it. ??
June 16, 2026 at 2:11 am
...
I don't want my life depending on approximations because it saved the doc 10 minutes of time, and I don't think AI should be let off with the excuse that humans make mistakes too.
Fair point, and as more of us AI, we need some accountability from the model trainers.
June 16, 2026 at 2:14 am
WOW Adrienne, what a story you relayed that happened to you!! And it makes me think, the last time I saw my doctor he did the same thing, in that he asked me if I would allow an AI to listen to our conversation, to take notes. I said yes, but your story shows me that I must be much more cautious!! What happened to you doesn't give me confidence that AI is better than humans at assisting with diagnosing. There is another problem, though. During that meeting the doctor told me that the medical company he works for is experiencing layoffs as well. Last week they had a cut in staff. So, I'm sure the temptation to use AI has increased just because they're cutting staff. This article, your experience, and staff cuts at medical providers suggest to me that misdiagnosis of medical conditions is going to go up. People are playing with fire without even seeing the flames.
What's scary about this is I worked for a doctor a long time ago. He would dictate notes after a session, and someone would transcribe them, but I have no idea how much he reviewed the transcription.
My wife later worked for Dragon, and even as doctor's dictated notes, Dragon software (voice recco, not AI) was used to transcribe the notes. Doctors were (I believe) responsible, but I wonder if that move from that to AI listening is abdicating more responsibility.
A good reason to record sessions yourself.
June 16, 2026 at 2:15 am
Indeed AI creators don't know how it works. They try to mimic structures already observed in the brain, and when a given threshold in layers and complexity of reached, then it starts to work. Then it comes the training phase, and here it's what liability could happen. But as long as there was no bad intention or bias in the training, I understand the doctors should interpret the results before taking decisions.
Yes and no. There is a little of a black box, but I believe that a lot of work has been done to understand how the models work. Now, that's different than the people training models understanding how the model works.
June 16, 2026 at 2:17 am
I believe that the issue here is data vs. information, then actions taken upon the information.
...
The point is, everyone involved in an incident is responsible for their portion of the cause. AI is obviously part of the cause of an incident where someone triggering the incident used the AI. I'm looking at AI as a service because that's what it is, although in cases where AI is also the program (i.e. the one performing the action) like Claude Code, Microsoft Agent Framework, etc.
Me head hurts here thinking about this. The ease of lawsuits and deep pockets distort this a bit, but it is a mess in deciding and apportioning out responsibility.
Viewing 9 posts - 1 through 9 (of 9 total)
You must be logged in to reply to this topic. Login to reply