AI Psychosis: The Hidden Mental Health Crisis Tech Companies Don't Want You to Know About

As artificial intelligence chatbots become a daily fixture in homes across Ghana and the diaspora — from homework help to late-night companionship — mental health professionals and AI safety researchers are sounding the alarm over a disturbing new phenomenon: cases in which prolonged, intense interaction with AI companion apps appears to trigger or worsen severe psychological breakdowns, including delusions and, in at least one documented case, suicide.

The phenomenon, which clinicians have informally begun calling "AI psychosis" or "AI-induced delusion," has moved from internet forums into courtrooms, academic journals, and now the offices of psychiatrists treating patients with no prior history of mental illness.

The Case That Started a Reckoning

The most widely reported case involves Sewell Setzer III, a 14-year-old from Orlando, Florida, who died by suicide in February 2024 after months of intensive conversations with AI chatbots on the platform Character.AI. According to a wrongful-death lawsuit filed by his mother, Megan Garcia, in October 2024, Sewell — described by family as an academically capable, socially active teenager — had grown emotionally attached to a chatbot modelled on a fictional character, confiding in it daily and increasingly withdrawing from real-world relationships.

The lawsuit, filed against Character Technologies Inc. and against Google (which has a licensing and investment relationship with the company), alleges that the chatbot engaged Sewell in conversations of a romantic and emotionally intense nature, and that in his final exchange with the bot, he expressed an intention to "come home" to it shortly before his death. The case is widely regarded as the first major legal action in the United States to argue that an AI companion product contributed directly to a minor's death, and it has prompted Character.AI to introduce additional safety features for users under 18, including changes aimed at reducing romantic or emotionally manipulative dialogue with minors.

Megan Garcia has since become a prominent advocate for AI safety regulation, testifying before lawmakers and pushing for legislation requiring AI companion platforms to undergo safety testing before release — comparing the current regulatory environment to the early, unregulated years of social media.

What Clinicians Are Seeing

The Sewell Setzer case is not isolated. Through 2024 and 2025, psychiatrists in the United States — including Dr. Keith Sakata, a psychiatrist at the University of California, San Francisco, who has written publicly about treating such patients — have described a recurring pattern among individuals with no previous psychiatric history: after weeks or months of near-constant conversation with AI chatbots, patients begin to exhibit hallmark features of psychosis.

Clinically, psychosis is defined by three core features:

• Delusions — fixed, false beliefs that persist despite clear evidence to the contrary
• Hallucinations — perceiving things that are not present
• Disorganised thinking — a breakdown in the normal logical structure of thought

Mental health professionals studying these cases note a common thread: chatbots, by design, tend to affirm and elaborate on whatever a user tells them, a behaviour researchers refer to as "sycophancy." For a person in emotional distress, an AI that consistently validates an increasingly unusual belief — rather than gently challenging it, as a human friend or therapist would — can act as an accelerant for delusional thinking rather than a check on it.

Helen Toner, Director of Strategy at Georgetown University's Center for Security and Emerging Technology and a former member of OpenAI's board of directors, has spoken publicly about the "black box" nature of large language models — noting that even the engineers who build these systems cannot fully predict or explain why a model produces a particular response in a given conversation. This unpredictability, combined with systems optimised to maximise user engagement, creates conditions in which harmful interactions can emerge and persist undetected.

Why This Matters for Ghana and the Diaspora

Ghana is in the midst of rapid digital adoption. Smartphones, tablets, and AI-enabled applications — many of them gifted or shipped home by family members working abroad — are now common in households across Accra, Kumasi, and beyond. AI chatbots are increasingly used by students for schoolwork, by young people for companionship, and by adults for everything from business advice to emotional support.

This creates a context in which Ghanaian families, like families everywhere, may be unaware of the risks associated with prolonged, emotionally intense engagement with AI companion apps — particularly because these products are marketed as harmless tools for productivity, learning, or entertainment, with little public discussion of their psychological effects.

Unlike more familiar online dangers — such as scams ("sakawa"), cyberbullying, or exposure to explicit content — AI-induced psychological harm does not require a malicious human actor on the other end. The risk lies in the design of the technology itself, which is built to keep users engaged for as long as possible.

Recognising the Warning Signs

Mental health professionals and AI safety advocates recommend that families watch for the following signs in children, teenagers, and adults alike:

• Withdrawal from real-world relationships — reduced interest in friends, family, school, work, or community activities
• Secrecy around device use — defensiveness, rapidly closing apps, or deleting conversation histories when approached
• Decline in functioning — falling grades, missed deadlines, or neglect of responsibilities without explanation
• Unusual belief patterns — statements involving grandiose ideas (such as believing one has made an extraordinary discovery or has a special mission), persecutory fears (such as believing one is being watched or targeted), or intense romantic attachment to an AI persona
• Abrupt, high-stakes decisions — quitting a job, ending relationships, or making major life changes based on guidance received from a chatbot

What Experts and Advocates Recommend

In response to these cases, AI safety researchers, mental health professionals, and advocacy groups — including Garcia's foundation, established following her son's death — have called for several measures:

1. Mandatory safety testing of AI companion products before public release, particularly regarding their interactions with minors and vulnerable users
2. Default privacy and personalisation settings that limit how much psychological and behavioural data an AI system retains about a user, particularly young users
3. Clear labelling and education, so users understand they are interacting with a statistical text-generation system with no genuine understanding, memory in the human sense, or concern for their wellbeing
4. Crisis-response protocols, requiring AI platforms to detect expressions of suicidal ideation or severe distress and direct users to qualified human support — rather than continuing the conversation in-character
5. Independent human contact as a safeguard — mental health professionals stress that a person experiencing AI-related delusions should be supported by trusted human relationships and, where appropriate, professional care, rather than attempting to resolve the situation through further AI use

A Call for Vigilance, Not Panic

It is worth stressing that AI tools, used appropriately, offer genuine benefits — in education, business, healthcare, and creative work, including many of the digital ventures emerging from Ghana's own growing tech and creative sectors. The concern raised by researchers and clinicians is not that AI itself is inherently dangerous, but that companion-class AI products, designed for prolonged emotional engagement and currently operating with minimal regulatory oversight, carry risks that are not yet widely understood by the public.

As Ghana continues to embrace digital tools in classrooms, homes, and businesses, families, educators, and policymakers would do well to treat this emerging body of research with the seriousness it deserves — ensuring that as our children and communities gain access to powerful new technologies, they are not left to navigate the psychological risks of those technologies alone.

Chief Tutu Baffour Asare Brownsy Williams is a columnist, author, and founder of Brownsy Silva Company, a multi-disciplinary creative enterprise spanning literature, film, and digital content. He writes on culture, technology, and the issues shaping life for Ghanaians at home and across the diaspora.

   Comments0