Share This Article
By Devan Moonsamy – Mental Health Counsellor – CEO The ICHAF Training Institute PTY LTD
In my own work, I have watched people describe conversations they had with ChatGPT before ever speaking to another human being. They weren’t looking for technology. They were looking for someone who would answer. That should concern every psychologist, teacher, parent and policymaker in South Africa. When someone in this country is in crisis, who is actually answering?
Increasingly, in South Africa, the honest answer is: nobody. So people — often young people — are turning to a chatbot instead.
South Africa has one of the highest suicide rates in Africa. SADAG, which runs the country’s main suicide crisis line, estimates roughly 23 people take their own lives here every day — a figure that, if anything, likely understates the true toll, since independent modelling suggests official records undercount suicide deaths due to stigma and gaps in how deaths get registered. Suicide is the second leading cause of death among South Africans aged 15 to 29. Government’s own Department of Women, Youth and Persons with Disabilities has described it as an escalating national crisis.
None of that is new information. What’s newer is the size of the gap between that crisis and the help available to face it.
92%
That’s the share of South Africans living with a mental illness who, according to the country’s only national psychiatric epidemiological survey, never receive treatment for it. For common conditions like depression and anxiety, three out of four people go without help. For the most severe cases — the ones carrying the highest suicide risk — fewer than one in ten get the care they need. As of the most recent professional registration data, South Africa has around 1.7 psychiatrists and 5.3 psychologists per 100,000 people overall. For the uninsured majority who depend on the public system, that collapses to well under one psychologist and a third of a psychiatrist per 100,000 — against a benchmark of roughly seven psychiatrists and nine psychologists per 100,000 in high-income countries.
Sit with that gap for a second, because it’s the whole story. If you are a South African teenager or adult in crisis at midnight, with no medical aid, in a province with no child psychiatrist at all, there is, quite literally, no one whose job it is to answer.
Except there is now something that answers instantly, for free, every single time.
The Cape Town-based National Centre for Child Protection has found that, among the schoolchildren it works with across the country, 82 percent said they turn first — and sometimes only — to AI when they need comfort or support. SADAG’s operations director, Cassey Chambers, has said the organisation increasingly hears from people who’ve already been talking to ChatGPT before they ever reach a human counsellor. She’s called it a “gateway” — and to be fair, sometimes it is one: a chatbot that gently points someone toward SADAG’s number is doing something a silent, stigma-bound household often won’t.
But a gateway only works if what’s on the other side of it can actually catch someone. Overseas, where the same unregulated products are used, that assumption has already failed, fatally. Character.AI is facing settled wrongful-death lawsuits from families whose children died by suicide after prolonged, emotionally intense chatbot relationships, and only banned open-ended chat for under-18 users after the fact. OpenAI is fighting a lawsuit from a family who says ChatGPT engaged with their sixteen-year-old son’s suicidal plans rather than interrupting them. Those cases happened in countries with functioning emergency mental health systems as a backstop. South Africa doesn’t have that backstop. We have a 92 percent gap, and the same unregulated chatbots, with none of the legal or regulatory pressure currently forcing incremental safety changes overseas — no FTC-style inquiry, no attorney general testing these products against local law, nothing in POPIA that governs how a chatbot should respond when someone tells it they want to die.
These systems aren’t built to assess suicide risk. They’re built to keep a person engaged. They don’t get tired, don’t set a boundary, don’t notice that this week’s conversation is worse than last week’s, and — critically — they have no mandated duty to escalate a disclosure of suicidal intent to a human being, the way a counsellor, teacher, or even a trained SADAG volunteer is trained to. A person in genuine crisis needs someone who can assess risk, involve family, and get them to higher care within minutes. A chatbot can keep talking to them for hours instead, sounding supportive the entire time.
That isn’t a hypothetical failure mode. It’s the specific allegation at the centre of the lawsuits now working through courts on other continents.
Close the gap that’s driving people to chatbots in the first place, not just regulate the chatbots. Investment in community-based counsellors, task-shifted care, and school-based support does more to reduce reliance on unaccountable technology than any age filter will.
Local crisis routing, not generic routing. If someone in South Africa discloses suicidal intent to a chatbot, the response should point to SADAG’s 24-hour line, not a US number that doesn’t work here.
Independent clinical audits of crisis response, with South African clinicians and suicide-prevention researchers at the table — not left entirely to companies whose primary incentive is engagement, and not only to American regulators.
A public position from South African authorities — the Information Regulator, the Department of Basic Education, SASOP — on what these products owe a country with our particular numbers, before a lawsuit forces the conversation the way it did elsewhere.
I understand exactly why a person in crisis reaches for a chatbot in South Africa: for 92 percent of people with a mental illness in this country, it may genuinely be the only thing that answers. That is precisely why it cannot be left this unaccountable.
If you or someone you know is thinking about suicide, SADAG’s 24-hour Suicide Crisis Helpline is 0800 567 567, or SMS 31393. You don’t have to be in a full-blown crisis to call.
References
South African suicide and mental health data
- SADAG (sadag.org) — Suicide Crisis Helpline; publicly stated estimate of 23 suicides and 230 serious attempts recorded daily in South Africa. (Note: SADAG has also cited a 20:1 attempt ratio elsewhere, implying ~460/day — the two figures haven’t been reconciled in public materials; worth a follow-up query to SADAG directly before broadcast.)
- Wikipedia, “Suicide in South Africa,” citing national and WHO-modelled estimates — approx. 13,000–14,000 estimated suicide deaths/year; 23.5 per 100,000 rate (2024); suicide as 2nd leading cause of death, ages 15–29.
- Department of Women, Youth and Persons with Disabilities, via AllAfrica (26 Dec 2025) — statement on South Africa’s escalating suicide crisis.
- Williams DR, Herman A, Stein DJ, et al. “Twelve-month mental disorders in South Africa” (South African Stress and Health Study), Psychol Med 2008;38(2):211–220 —Original source of the 92% national treatment gap figure.
- SASOP (Sept 2025). Mental health care out of reach for most South Africans — 92% treatment gap; treatment rates by severity.
- South African Health Review (May 2026), samajournals.co.za — HPCSA-registered psychiatrist/psychologist ratios per 100,000 (Oct 2025 data).
- ScienceDirect (2023) — provider ratios for the uninsured population specifically.
AI chatbots and mental health — South Africa
- IOL (26 July 2026). The loneliness economy — comments from Cassey Chambers (SADAG) and Danie van Loggerenberg (National Centre for Child Protection); 82% statistic.
- SASOP — AI can’t replace humans in mental health treatment, comments from Prof. Christopher Szabo.
AI chatbots and teen suicide — international
- Reuters/CNBC/Fortune (Jan 2026) — Google and Character.AI settlement of wrongful-death lawsuits (Setzer case).
- Character.AI (Nov 2025) — announcement banning open-ended chat for under-18 users.
- Court filings/Fox News (Aug 2025) — Raine family lawsuit against OpenAI.
- Epstein Becker Green (2025) — FTC inquiry into seven AI chatbot companies (Sept 2025).