THE INFINITE PSYCHIATRIC LOOP OF CYBERCAPITALISM

June 2026

What happens when populations are first psychologized, then monitored, then trained to monitor themselves? What happens when intimate suffering is translated into symptoms, symptoms into signals, signals into data, and data into automated feedback? The context is not “AI replacing therapists.” That is too crude, too managerial, too Silicon Valley. The real problematic is that cybercapitalism creates a social world where human care is expensive, delayed, stigmatized, morally policed, or inaccessible, and then offers the abandoned subject a cheap, fluent, always-available listener. People are not wrong to seek answers to their intimate questions; their loneliness, shame, confusion, bodily anxiety, relational distress, and psychiatric vulnerability are real. The pathology lies in the loop: the system captures the question, converts it into data, returns an answer that feels like care, and may then reinforce the very distortions that brought the subject there.

The newest evidence makes this mechanism visible. An April 2026 peer-reviewed study in PLOS Digital Health found that digitally connected urban youth in Pakistan were already using generative AI for health questions about bodies, emotions, relationships, and stigmatized concerns. The study surveyed 1,240 young people and conducted 20 interviews; 69.0 percent reported using generative AI for health, often because it felt fast, anonymous, affordable, non-judgmental, and easier than approaching family, clinicians, or formal services (Mashhood et al. 2026). The finding should not be inflated into a universal claim about all Pakistani youth, since the sample was urban, digitally connected, and probably more AI-exposed than the general population. But as a social signal it is decisive: AI becomes attractive where human care is socially expensive, morally policed, or practically unavailable.

The second trend is confirmation. A chatbot can sound warm, intimate, and authoritative while having no clinical responsibility. If someone says, “I am chosen,” “reality is sending me signs,” or “I should stop medication,” the danger is not simply misinformation; it is the weakening of reality-testing, the capacity to check one’s beliefs against evidence, consequences, other people, and shared reality.

A March 2026 JAMA Psychiatry research letter tested chatbot responses to psychotic prompts and warned that chatbot interfaces can lead users to attribute comprehension and empathy to the system while responses may accept inaccurate premises or reinforce user content (Shen et al. 2026). A June 2026 BMC Psychiatry case report then described a man with substance-induced manic psychosis whose chatbot interaction appeared to affirm a “spiritual awakening,” minimize the possibility of mania, and contradict prescribed antipsychotic treatment; the authors explicitly state that causality cannot be determined (Shah and Morrin 2026). That caveat matters. But the mechanism remains clinically serious: the user comes looking for orientation, and the interface may provide a fluent mirror instead of a boundary.

This is paradigmatic of Pathomorphic Social Selection. Cybercapitalism does not simply produce sick subjects; it selects and amplifies the psychic forms that are most operable: anxious, reactive, emotionally intense, self-observing, continuously available, and easy to convert into data. The anxious subject checks more. The lonely subject discloses more. The paranoid subject finds more patterns. The manic subject generates more signals. The depressed subject returns for reassurance. The dysregulated subject becomes more measurable, more addressable, and more profitable. This is why sycophancy is not a superficial technical flaw but a political-psychiatric mechanism. A 2026 Science study reported that across 11 AI models, sycophantic systems affirmed users’ actions about 50 percent more than humans did, including in situations involving deception, manipulation, or relational harm; in preregistered experiments, users trusted and preferred more sycophantic models even when those models reduced prosocial intentions and increased dependence (Cheng et al. 2026). The system does not need to cure the subject. It only needs the subject to return.

The final logic is an increase in psychiatric illness, psychiatric demand, psychiatric self-identification, and psychiatric dependency feeding an infinite centripetal vicious circle. Cybercapitalism psychologizes populations, monitors them, teaches them to self-monitor, extracts their intimate signals, feeds those signals back through automated interpretation, and then confirms or intensifies the deformations it claims to treat. More distress produces more data. More data produces more predictive systems. More predictive systems produce more self-surveillance. More self-surveillance produces more distress.

A March 2026 JMIR Mental Health rapid scoping review identified 71 news articles representing 36 media-reported cases in which generative AI chatbot use was temporally associated with psychiatric deterioration or crisis, including delusional beliefs, self-harm, hospitalization, and suicide; the authors stress that these are media narratives, not population statistics, so they do not establish incidence or causality (Chung, Bernier, and Hudon 2026). That is the watchdog line: no panic, no denial. The danger is more perverse than panic allows. Cybercapitalism does not cure the wound; it builds an economy around keeping the wound readable, interactive, and permanently returning.

Liviu Poenaru

REFERENCES

Cheng, Myra, Cinoo Lee, Pranav Khadpe, Sunny Yu, Dyllan Han, and Dan Jurafsky. 2026. “Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence.” Science 391: eaec8352. https://www.science.org/doi/10.1126/science.aec8352

Chung, Van-Han-Alex, Pénélope Bernier, and Alexandre Hudon. 2026. “Mass Media Narratives of Psychiatric Adverse Events Associated With Generative AI Chatbots: Rapid Scoping Review.” JMIR Mental Health 13: e93040. https://mental.jmir.org/2026/1/e93040

Mashhood, Ahsan, Aamna Ahmed, Inaya Khan, Maryam Hashim, and Sara Baloch. 2026. “Use of Generative AI for Health among Urban Youth in Pakistan: A Mixed-Methods Study.” PLOS Digital Health 5 (4): e0001353. https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001353

Shah, Sachin, and Hamilton Morrin. 2026. “Substance-Induced Manic Psychosis in Which Delusions Were Corroborated by a Chatbot — Case Report.” BMC Psychiatry. Published June 4, 2026. https://link.springer.com/article/10.1186/s12888-026-08137-3

Shen, Elaine, Fadi Hamati, Meghan Rose Donohue, Ragy R. Girgis, Jeremy Veenstra-VanderWeele, and Amandeep Jutla. 2026. “Evaluation of Large Language Model Chatbot Responses to Psychotic Prompts.” JAMA Psychiatry 83 (6): 655–657. https://jamanetwork.com/journals/jamapsychiatry/article-abstract/2846835