When health leaders from more than 100 countries sat down to define responsible AI, they reached a striking conclusion: the technology is rarely the hard part. On 1 September 2026, a World Health Organization forum reported that the future of AI in mental health and healthcare depends on the strength of the governance around it, not the sophistication of the tools (World Health Organization Regional Office for Europe, 2026). For South African psychologists weighing AI in their own work, that message lands close to home. It also shapes your ethical CPD priorities for the year ahead.
The global signal: governance, not gadgets
The WHO report drew on its first Knowledge Community on responsible AI in health, a structured online dialogue held in late 2025 (World Health Organization Regional Office for Europe, 2026). Participants came from 105 countries, and they agreed on a blunt point. The main obstacle to safe AI is not the model. Rather, it is the weakness of the institutions meant to govern it.
They named familiar culprits: fragmented and biased data, unclear accountability, and thin AI literacy among staff. Furthermore, they argued that ethics cannot be a box ticked once at launch. Instead, ethical review must run through the daily life of a health service, including the willingness to pause or drop a tool that cannot be governed safely.
Why AI in mental health lives or dies on oversight
Mental health raises the stakes further. The data is intimate, the risk judgments are subtle, and a wrong call can cost a life. Consequently, an AI tool that screens for depression or flags suicide risk needs more scrutiny, not less.
The oversight trap
Oversight only works when clinicians can understand and challenge what a tool suggests. However, many systems stay opaque, and busy practitioners rarely get time to interrogate them. As a result, human oversight can shrink into rubber-stamping, exactly what the WHO forum warned against (World Health Organization Regional Office for Europe, 2026).
South Africa is hearing the same warning
This is not only a European conversation. In June 2026, the South African Medical Association’s AI Task Team published its own appraisal of the HPCSA’s Booklet 20, the first local ethical framework for AI in healthcare (Mahomed et al., 2026). Its verdict rhymed with the WHO’s.
The Task Team praised Booklet 20 as a foundation, yet it flagged gaps in accountability, human oversight, disclosure and data sovereignty. Moreover, it called for practical safeguards — risk tiers, consent templates and bias checks — grounded in the ubuntu principle of relational accountability. In short, two bodies on two continents reached the same warning in the same season: governance is lagging the technology.
Understanding this shift is fast becoming part of good practice and of your CPD. PsyCampus’s HPCSA-accredited CPD packages include ethics CEUs that help you apply these standards with confidence.
What strong governance looks like for AI in mental health
Governance can sound abstract, so it helps to make it concrete. At a practice level, strong governance means knowing your tools, protecting your clients’ data, and keeping a human firmly in the loop.
Questions worth asking before you adopt a tool
Before you bring any AI tool into your practice, ask a few plain questions:
- Who built it, and on what data was it trained?
- Can it explain its outputs in terms you can defend clinically?
- Where is client data stored, and does it comply with POPIA?
- What happens when it is wrong, and who is accountable?
If a vendor cannot answer these clearly, that silence is itself an answer. Therefore, treat unexplained systems with the caution that both the WHO forum and the SAMA Task Team recommend.
Practical steps for your practice
You can act now, without waiting for regulators to catch up. Specifically, small habits build the culture of oversight that both reports describe. Start by documenting the AI tools you use and reviewing their outputs before they reach a client file.
Additionally, tell clients when AI supports their care, and keep your consent language plain. Finally, treat AI ethics as ongoing learning rather than a one-off box to tick, because the standards will keep moving, and staying current protects both your clients and your registration.
Want to stay ahead of the governance curve? Browse PsyCampus’s 2026 HPCSA-accredited CPD packages with dedicated ethics CEUs, and see whether you qualify for a graduate discount of up to 40%. For the South African regulatory detail, read our companion post on what the HPCSA AI guidelines mean for psychologists.
References
- Health Professions Council of South Africa. (2025). Ethical guidelines on the use of artificial intelligence in healthcare (Booklet 20). HPCSA. Retrieved from https://www.hpcsa-blogs.co.za/updated-ethical-guidelines/
- Mahomed, S., Ncube, M. V., Dhai, A., Janneker, W., & Nodikida, M. (2026). Ethical governance of artificial intelligence in healthcare: Critical reflections of the AI Task Team of the South African Medical Association on the Health Professions Council of South Africa guidelines on the ethical use of AI in healthcare (Booklet 20). South African Medical Journal, 116(5), e5072. https://doi.org/10.7196/SAMJ.2026.v116i5.5072
- World Health Organization Regional Office for Europe. (2026). Progress on AI in health should be determined by strength of governance, WHO forum urges. Retrieved from https://www.who.int/europe/news/item/01-09-2026-progress-on-ai-in-health-should-be-determined-by-strength-of-governance–who-forum-urges