Home » Insights » HPCSA AI Guidelines: Why It Matters

HPCSA AI Guidelines: Why It Matters

Artificial intelligence is arriving in South African consulting rooms faster than the rules meant to govern it. In late 2025, the Health Professions Council of South Africa released Booklet 20, and the resulting HPCSA AI guidelines became the first ethical framework for AI use across every registered profession. Furthermore, in June 2026 the South African Medical Association’s AI Task Team published a detailed appraisal of that framework, and it flagged serious gaps. If you already lean on AI for note-taking, screening or assessment, these developments shape your ethical CPD obligations directly.

Why the HPCSA AI guidelines matter now

AI tools no longer sit on the edges of clinical work. Instead, they draft session notes, flag risk in intake questionnaires, and power the chatbots some clients meet before they ever reach you. Consequently, the HPCSA needed a clear position, and Booklet 20 is that first attempt (Health Professions Council of South Africa, 2025).

The framework sets out broad ethical principles: patient safety, confidentiality, informed consent, and professional accountability. Moreover, it reminds practitioners that a registered professional — not an algorithm — remains responsible for every clinical decision. That principle sounds obvious. In practice, though, it raises hard questions the booklet does not fully answer.

What the SAMA Task Team found missing

The SAMA AI Task Team welcomed Booklet 20 as an important foundation (Mahomed et al., 2026). However, the team argued that principles on their own cannot guide safe use. Specifically, the appraisal identified operational gaps across human oversight, disclosure, data sovereignty, equity, accountability, and risk categorisation.

Principles without implementation tools

The Task Team’s central concern is practical. Booklet 20 tells practitioners what to value, yet it offers few tools for doing so. Therefore, the authors call for concrete mechanisms — risk-tiered safeguards, structured consent templates, and context-appropriate standards for transparency, explainability and bias mitigation (Mahomed et al., 2026).

Additionally, they anchor these fixes in ubuntu, South Africa’s communitarian ethic. In that view, accountability is relational, and communities deserve a say in how AI enters their care. As a result, the appraisal reads less like a rejection and more like a blueprint for the next version of the guidelines.

Accountability and human oversight gaps

For psychologists, the accountability gap matters most. When an AI screening tool misjudges suicide risk, who answers for it? Booklet 20 places responsibility on the practitioner, yet it says little about the vendors, data and systems that shape a tool’s behaviour (Mahomed et al., 2026).

Human oversight faces the same problem. The guidelines expect clinicians to supervise AI outputs, but meaningful oversight needs time, training and explainable systems. Without those supports, oversight becomes a signature on a form rather than genuine clinical judgment.

Ethics sits at the heart of every HPCSA CPD cycle, and AI is quickly becoming part of that conversation. You can build these skills through PsyCampus’s HPCSA-accredited CPD packages, which include dedicated ethics CEUs for 2026.

What the HPCSA AI guidelines mean for your daily practice

So how should you respond right now? First, treat every AI output as a draft, not a verdict. You remain the clinician of record, and your notes should reflect your own reasoning.

Disclosure and consent

Tell clients when AI plays a role in their care. Booklet 20 and the Task Team both stress disclosure, and clear consent protects trust as much as it protects you (Health Professions Council of South Africa, 2025; Mahomed et al., 2026). A short, plain-language explanation usually does the job.

Data protection under POPIA

Client data feeding an AI tool still falls under the Protection of Personal Information Act. Therefore, check where that data travels, who stores it, and whether it leaves the country. Data sovereignty is one of the gaps the Task Team highlighted, so caution here is well placed.

How to strengthen your ethical AI practice

You do not need to wait for a perfect framework to practise well. Instead, you can adopt the safeguards the Task Team recommends, one step at a time. Practical habits that align with the guidance include:

  • Keeping a written record of which AI tools you use and why.
  • Reviewing every AI-generated note or score before it enters a client file.
  • Explaining AI involvement to clients in clear, plain language.
  • Storing client data with POPIA-compliant, ideally local, providers.

Finally, keep learning. The rules will keep changing, and a World Health Organization forum reached the same conclusion in September 2026 — governance, not technology, decides whether AI helps or harms (World Health Organization Regional Office for Europe, 2026). Staying current is itself an ethical duty.

Ready to turn these principles into confident practice? Explore PsyCampus’s 2026 HPCSA-accredited CPD packages with built-in ethics CEUs, and check whether you qualify for a recent graduate discount of up to 40%. For the wider global picture, read our companion post on why AI in mental health depends on governance.

References

Shopping Basket
Scroll to Top