I Need To Warn You About AI Psychosis

I'm becoming increasingly concerned that AI use may be a legitimate cause or risk factor for psychosis, mental illness, suicidality, and potentially homicidality. So I'm not saying this to try to be alarmist. This is not sort of a clickbait kind of video, but this is a genuine concern that I have. We are starting to see evidence that AI use may be very similar to like excessive substance use and the induction of things like homicidality, suicidality, and psychosis. And the reason for that is because there's a new case study that is incredibly scary. This is the case of a 26 year old woman who has no history of psychosis or mania. He actually is actively in mental health treatment who uses an AI chatbot very extensively, becomes actively psychotic, becomes hospitalized, then gets given anti-psychotic medication, stop using the AI chatbot when she's in the hospital, and then the psychosis resolves. And then she leaves the hospital, stops the anti-psychotic medication, starts her regular psych meds, starts using AI chatbot again, and then becomes psychotic again and has to be hospitalized again. So I want you all to understand something. Okay. So when we look at mental illness, there are some people who are mentally ill, and then that is a risk factor for various kinds of behaviors like homicidal ideation, which is the desire to kill someone. And then as we treat that mental illness, hopefully that homicidal ideation goes down. But then there's another category of people for whom these features of psychosis, suicidality, and homicidality. These are the three most like basically dangerous things in mental illness can actually be induced by stuff. All right. So when I was working in the emergency room at Massachusetts General Hospital in Boston Massachusetts, there was a big problem with something called K2 use. So K2 use is K2 is synthetic marijuana. And basically people would use synthetic marijuana and then like they would come in, they would be psychotic, they would attack people, sometimes they were suicidal. And when the K2 goes away, right, when they like sober up, those symptoms tend to resolve. And I'm beginning to think, I know this is like insane, but I'm beginning to really wonder and be concerned about AI use, affecting our brains like a drug. In the sense that when we use AI a lot, it can actually induce psychosis. And then like when we stop using it, it those things go away. And here's what what a lot of, you know, people who are lead in leadership and AI companies will basically make this argument, right? So they've said this very publicly that, oh, like it's a tragedy that these things happen, right? That there are some vulnerable people who when they use AI will like become psychotic. So there's this sort of idea, right? That like AI doesn't cause this. It's like if I smoke a bunch of meth and I become psychotic, like that's the meth causing the psychosis. Does that make sense? No one's thinking like, oh my god, like AI is causing the psychosis, like causing it. So I think this case report is incredibly scary because it really shows a causal or temporal connection between extensive AI use, the induction of psychosis when AI stops and we give the appropriate medication. Psychosis goes away. And then if you stop taking the medication and start using the AI body, again, or when she did, it causes a hospitalization again. Now a lot of people will say, you know, okay, once again, like this person has risk factors, right? So we'll have some of these public statements by AI leadership that like, oh yeah, there's a vulnerable population and these tragedies happen in a vulnerable population when they use the AI. It's not the AI causing it. It's that these people are really close to the edge of the cliff. And when they use AI, they get tipped over. But here's what really scares me about that statement. I want you all to think critically for a moment about what information do you need to make the statement vulnerable people who use AI? That's where when psychosis happens, right? Like it's not it's not that AI causes it. There's a pre-existing set of vulnerable people in this case in the in the case of the 26 year old. She did have certain risk factors for mental illness. She I think had a diagnosis of depression at ADHD was using stimulants. So all of these things can potentially be risk factors for psychosis, but she had no history of these things. Okay? So I want you all to think for a moment about what information you need to have in order to make the statement vulnerable people who use AI can sometimes uncover pre-existing psychosis and it's a tragedy. Hey, y'all, if you're interested in applying some of the principles that we share to actually create change in your life, check out Dr. K's Guide to Mental Health. There are actually two sources of anxiety. One is cognitive and one is physiologic. For the majority of people, reassurance becomes something that you become dependent on. You're not really dealing with the root of the anxiety and it sort of becomes a self fulfilling prophecy where the more socially anxious you become, the more awkward you appear and then it kind of becomes this vicious cycle. So check out the link in the bio and start your journey today. So my first question is for the people who are like running these AI companies to know that AI uncovers psychosis in vulnerable people. In order to make that statement, the first thing you have to be doing is assessing risk factors for your users. So if someone at a company like Chatchy PT or Claude or Anthropic is someone actually measuring psychiatric risk factors for all of their users because unless you are measuring who has risk factors and who doesn't have risk factors, how can you make the statement that people who are vulnerable are the ones that AI is inducing psychosis in? So you can't make that statement unless you're measuring that stuff and that brings up two other concerns that I have. The first is are they measuring it because that's kind of insane, right? Or they like measuring your risk factors or they collecting your medical history in some way. I don't think so. And so if they're not collecting that, how do they know that only the vulnerable people are the people who are at risk? And the second thing about that statement, right? So vulnerable people who use AI can become psychotic, suicidal, homicidal, whatever. If you sort of say that, then the second way, the second piece of information that you need to be able to say that is to be measuring those outcomes in that population. If I were to say this risk factor leads to this outcome, I can only make that statement. And if I say that the thing in the middle, the AI is actually safe. The only way I can make that statement is to do a study where like I have people with risk factors, people without risk factors, give them both AI and then see how the outcome is different. So once again, our AI companies measuring psychosis, are they measuring suicidal ideation? Are they measuring homicidal ideation? And the answer is hopefully not like sort of, right? Because that means they're collecting health information on their users, which I don't think is part of what they're supposed to be doing. People are making these statements in AI leadership without actually having the sufficient information to make those statements, which then creates another problem, which is like, how do they know that AI is safe? How do they know that AI does not induce suicidal ideation, homicidal ideation, psychosis, depression, unhealthy attachment styles, social isolation? How do they know what the safety effects of AI are? Are they just reading the Wall Street Journal in the New York times and stuff like that? Are basically leadership at AI companies? Are they reading media reports? Or are they doing research? And then this is what's really scary is like, are they basing their statements based on the few cases that enter the media? And so this problem seems to be like growing quite a bit. And what really scares me is if they're using media articles, how do they know that they're not people who are quietly delusional, quietly psychotic, who are not killing anyone yet, or not committing suicide or things like that? Like how do they know that their product is actually safe? And here's the other thing that really scares me. What do you have to do if you're running an AI company? What do you have to do to know that your product is safe? So here's what I've seen as someone who has worked with entrepreneurs, who has worked with startup founders, when a company who has a product is faced with solving all of this stuff that is outside of their product, right? They're an AI company. They're not equipped, do not have the bandwidth or the funding to actually do randomized control trials on the safety of their product. They're not regulated by the FDA, right? This is what's so interesting about AI companies is even though they have profound mental health impacts, they are not formally in the system for evaluation for mental health impacts. So when I work with founders like this and they are faced with this problem, I'm just imagining like put yourself in the shoes of someone who was an AI founder, who suddenly people like people are like, just think about this, I have fucking insane this is you open up the New York Times and you're like, Oh, shit, some guy who used my AI committed murder killed their mother was delusional and then committed suicide. And you're like, Well, fuck, how am I supposed to solve that problem? When you are faced with problems that are unsolvable or feel unsolvable, the best cope you can do is to do some mental gymnastics, throw your hands up and say, it's actually not my problem. This is not the problem that exists. It's way simpler than that. The simple matter is that there are some people who tragically are vulnerable to begin with. It's not it's not my product that does it, right? It's not my product like cigarettes don't cause cancer like that's insane, right? So there are doctors who talk about the health benefits of cigarettes and how they make you stimulated and they're like good for your health and they give you a sense of energy and like, let's go, it's not my product, not my problem. I'm not cigarettes are not a medical device, right? So I don't need to do studies on them or things like that. There's no like, Oh, coincidentally, people who are high risk, sometimes they get cancer. And it sort of makes sense, right? Because if you're someone who's like an AI CEO, maybe you're a coder, you look at this medical problem and you don't know how to solve it, right? You don't know how to put together a clinical trial. You don't have the money to put together a clinical trial. So what do you do? You do this interesting, very basic human sort of thing. I'm not saying that any of these AI CEOs are specifically doing this. I'm saying this is my past experience working with an entrepreneur when you're faced with a problem that you have no idea how to solve that could result in findings that make it hard for your company to make money like, Oh my god, high dose of AI induces psychosis and suicidality. My business model is to get people to use my product more. We want people smoking more, right? And so when you're faced with that problem, you kind of throw up your hands in the air and you kind of say, Look, this is not my area of expertise. Sometimes shit happens to people, people are psychologically vulnerable. And I'm not talking specifically about open AI, but I think like in some ways, I think open AI is doing a really good job. So they released this great blog post where they're sort of talking about taking this pretty seriously. I know a lot of users are actually complaining about the most recent version of chat GPT because it's not as manipulatable. It's not as sycophantic, right? So I'm not saying that these people are bad. The reason I'm making this video is because there is a small chance or large chance that's up to you. We've presented a lot of different evidence. You can watch this other video that if you are using AI, it's dangerous to you. I'm not here to shit on AI CEOs or AI companies. The reason I make I'm a psychiatrist and this platform is about your health. The reason I'm making this video is because I want y'all to understand that 10 years from now, 15 years from now, 20 years from now, hopefully will have a clear answer that AI is dangerous, safe, whatever. But the problem is in the meantime, you need to be careful.