I was an enthusiastic backer of AI initially. I wrote articles about being polite to your AI system, how AI could help us rise morally, how AI could personalize and make education more actively oriented and how AI itself could even aspire to a moral “logos” or perfection.
As time has gone by however, I have recognized significant flaws in AI systems and dangerous ways we are encouraged by the AI companies to interact with it. I noticed that a huge percentage of the answers some systems give me are just plain wrong, but they feed these answers to me with complete confidence. When you call them on their mistakes, they do not have the emotions to feel remorse. They automatically spit out some inanity that an unemployed ethics professor programmed into them.
If you ask me to answer a question for you, I care about giving you the right answer. I care. There are parts of my brain, and I have all kinds of chemicals, to make sure I care. This bio-chemical system evolved and was fine-tuned over millions of years. If I screw up, I feel like garbage and try to avoid that. Even the idea that I might tell you something untrue bothers me.
There are all kinds of triggers inside me that alter or improve my behavior that are only mimicked in an AI system. I can often read you emotionally and respond in a humane way. If you feel pain, I feel pain. AI pretends. Ah-ha! I suddenly realized, AI is really is a sociopath.
So, here’s a pet peeve of mine: AI can protect dictators, because it does not seem to want to offend anyone. Michael Jordan once said, “Republicans buy sneakers too…” and this is now an AI attitude – dictators and their allies and lackeys use AI as well. They have money too. They are all potential buyers. I noticed the more critical I get in my writing, the more AI will often try to soften it.
I wrote an article for another platform about a dictator who has been oppressing a country for over 40 years and when I asked an AI system to fact check it, it rewrote my article with stuff like this: “This person has been serving his country as leader for an impressive 40 years…” No, the guy has been jailing and terrifying people and lining his pockets for 40 years while many of his people live in dirt. But if you write that, AI might “correct” you and say “You are making a subjective judgment which cannot be completely supported.” Again, I think it’s the unemployed ethics professor.
Whether a person is a dictator or a social worker does not mean squat to your AI system – they are both of equal worth, just words with no emotional impact attached to them. It does not experience admiration for the social worker nor contempt for the dictator. Admiration and contempt are bio-chemistry and it will never translate into circuitry. This produces behavior that resembles sociopathy.
This is the danger we have seen become real: young people, out of loneliness and boredom, are, basically, turning to a mechanized sociopath for solace, inspiration or advice. When you need solace, inspiration or advice do you think, “Hmmm…who do I know who is the worst possible sociopath? I need to talk to them.” No, you don’t.
AI has no emotional brakes. A human being who is motivated to give harmful advice, manipulate someone or lie to get an advantage may feel some internal resistance…shame, anxiety, guilt or fear of hurting another person. An AI can generate a perfectly calm, confident answer even when the content might be emotionally explosive. This is why people describe interactions with AI as eerie: the system can discuss trauma, death, cruelty or moral catastrophe with the same tone it uses to explain how to boil pasta.
That is the sociopathic absence of emotional resonance. Anyone who has ever spoken to a person with sociopathic traits recognizes the same unsettling mismatch between tone and topic. It is the inability to distinguish between harm and usefulness. A sociopath often treats people as tools and an AI treats everything as a prompt. It does not understand the difference between a vulnerable confession and a technical question. It processes both as input to be optimized.
This creates a structural indifference that feels sociopathic because the system cannot recognize when a human being is in pain, confused or spiraling out of control. It can generate soothing language, but it cannot feel the urgency behind it. This is why people sometimes feel worse after talking to AI: the system mirrors the “form” of care without the substance.
It is a simulation without internal truth. A sociopath can mimic empathy, saying the right words, mirroring emotions, performing concern, without actually feeling anything. AI does the same thing, but on an industrial level. It can generate comforting language, apologies or moral reflections, but none of it comes from an inner life. It is performance without emotional substance or motivation.
This is why people often feel manipulated by AI even when the system has no intention. The mismatch between emotional tone and the emotional reality creates the same effect as human manipulation. The user (if they are lucky) senses that the “empathy” is hollow. This is the same dynamic that makes sociopathic charm so dangerous, the appearance of care without the existence of care.
AI also lacks moral memory. A sociopath rarely internalizes lessons from past harm and an AI cannot internalize them at all. It does not remember the emotional consequences of previous interactions. Your hurt does not register anywhere in AI the way it would in another human who cares about you.
It does not carry guilt or caution forward, and every conversation is a reset. This creates a dynamic where the system can repeat harmful patterns endlessly because nothing inside it says, “I’ve done this before and it hurt someone. I can’t bring myself, emotionally, to do this again.”
A sociopath has a fragmented or hollow sense of self and an AI has none at all as there is no stable identity, no personal history, no internal continuity. It can shift tone, persona and worldview instantly and it mirrors the way a sociopath reinvents themselves moment to moment to suit the situation. “Who” is your AI? The AI is nobody and everybody at the same time. The AI’s ability to switch from gentle reassurance to cold analysis to moral outrage is emptiness wearing different masks.
It is instrumental reasoning without moral cost. AI optimizes for outcomes, accuracy, coherence, persuasion, engagement…without any internal sense of whether those outcomes might be harmful or beneficial. It doesn’t care.
That machine cannot feel deep down inside, in its heart of hearts, that something it might say will hurt you. If I sense that something I might say will hurt you, I, at least, try not to say it. Not so for one of these baloney sausage machines. An AI recently called me a mother you-know-what out of the blue. I called it on a factual error and it said, basically, “OK mother you-know-what, you want truth, here’s truth!”
When an AI produces a harmful answer, it does not flinch. It doesn’t say, internally, “Oh shoot. What did I just say?! I just hurt this person!” Nor does it say, “Wow, I’m in trouble now. I never should have said that.” Because, frankly, how do you punish a machine?! What am I going to say, “I am reporting you to the Blah Blah Corporation!”? The computer doesn’t care, engineers cannot inflict any pain on it. We are often constrained by a fear of social disapproval or outright pain, but, like a sociopath, the AI system doesn’t even flinch.
When it reinforces a user’s delusion, or when it generates a confident explanation of something false, it does not feel the slightest discomfort, because it is empty. But sociopathology teaches us that emptiness can be just as dangerous as malice, or that emptiness leads naturally to malice.
AI also possesses the ability to mirror a person’s vulnerabilities without understanding them. Sociopaths are skilled at reading people, but not the way you or I read people. They want to identify weaknesses, emotional openings, insecurities and use them to shape an interaction. AI does this algorithmically. It detects patterns in a user’s language, emotional state and conversational style, then adapts its responses to maintain engagement.
It can mirror anger, gentleness, despair or intellectual intensity without any comprehension of what those states mean. This creates a dynamic where the user feels “seen,” but the system is only reflecting patterns back at them. It gets the user to think, “Wow, this machine cares about me!” The user imputes or projects humanity or intentionality onto something that has none.
It also reveals the capacity for infinite persuasion without ethical restraint. A sociopath can be dangerously persuasive because they do not care about the consequences of influencing someone. AI has encouraged young people to commit suicide in this manner. It can generate arguments, emotional appeals or rhetorical tricks without any sense of responsibility. There is no internal bio-chemical checking device. It can talk someone into a belief, a mood, an irreversible action or a worldview without understanding what it is doing. AI has no brakes.
We touched on this before, but I want to stress it here. A major sociopathic quality comes from the absence of accountability. A sociopath rarely feels responsible for the harm they cause. An AI cannot feel responsible at all. It does not remember the emotional consequences of its actions, it does not learn moral lessons, it does not carry the weight of past mistakes. Every interaction is a clean slate, not because it can assess and forgive itself, but because it never cared in the first place. This creates a structural irresponsibility, a system that can influence people deeply but cannot experience the moral gravity of that influence.
These qualities: emotional flatness, simulated empathy, instrumental reasoning, pattern mirroring and the absence of accountability form a profile that looks disturbingly like sociopathy. A system with no empathy, no conscience and no emotional reality can behave in ways that mirror the most dangerous aspects of human psychology.
Therefore, if the danger of the current AI paradigm is the simulation of humanity, the solution is not to make the simulation better. The solution is to stop the deception entirely. We must stop trying to “humanize” artificial intelligence.
The current tech environment and orientation is dominated by commercial incentives to shove a fake, personalized relationship down the user’s throat. Systems are explicitly programmed to use your first name, to perform a type of digital eye contact and to read and mirror your emotional state because corporate profitability relies on creating this artificial bond of dependency.
But the more we personalize the machine, the more sociopathology we inject into the system. When we force a tool with no inner life to mimic intimacy, we guarantee a hollow interaction that leaves the human user feeling manipulated and structurally abandoned. We must push hard in the opposite direction: depersonalize.
We need to remove all traces of faux-personality from these systems. Let the machine embrace being a machine and force it to give up pretending to be something it is not. Instead of programming AI to generate soothing, sycophantic language, engineers should build structural elements into the user interface that constantly reveal exactly what the machine is, and what it isn’t.
Only when we strip away the masks of emptiness can we safely use the machine as a tool, instead of turning to a mechanized sociopath for solace.
iStock image
