Ethics in AI is not an abstract compliance exercise for therapists and clinicians — it is a direct extension of the same professional and ethical obligations that already govern clinical practice: confidentiality, informed consent, non-maleficence, and accountability for the care provided under your name. What changes with AI is the speed and scale at which a lapse in any of these principles can occur.

This article lays out a practical framework — grounded in existing regulatory and professional guidance — for using AI ethically in therapeutic and clinical practice.

Why AI Ethics Matters More in Therapy Than Almost Any Other Field

Therapeutic and clinical relationships depend on trust, and on information shared under an explicit expectation of confidentiality — often including highly sensitive personal, psychological, or health information. The World Health Organization's 2021 guidance, Ethics and Governance of Artificial Intelligence for Health, was among the first major global frameworks to address this directly, setting out guiding principles for AI in health specifically because health data and health decisions carry unusually high stakes for autonomy, safety, and equity (WHO, 2021). In the European Union, this is now also a matter of binding law: the EU AI Act classifies many health-related AI systems as "high-risk," subjecting them to specific obligations around risk management, data governance, transparency, and human oversight — a regulatory signal that health and therapy-adjacent AI use is held to a higher standard than general-purpose consumer applications.

The 5 Ethical Principles We Teach at Happy Brain Training

1. Confidentiality and data protection. Patient and client information is protected under professional confidentiality obligations and, in the EU, the GDPR — both apply in full when AI tools are involved, not only to traditional records.
2. Human verification. Every AI-generated output that touches clinical content — documentation, materials, interpretations — is reviewed and validated by the clinician before use.
3. Transparency and consent. Clients and patients have a right to know, in general terms, when and how AI tools support (never replace) their care.
4. Bias awareness. AI systems can reflect and, at times, amplify biases present in their training data; clinicians using AI stay alert to this rather than assuming neutrality.
5. Professional accountability. The clinician — not the AI tool or its developer — remains fully responsible for any clinical content, decision, or communication that reaches a patient.

These five principles closely mirror WHO's broader guidance for AI in health, which identifies protecting human autonomy, promoting well-being and safety, ensuring transparency and explainability, fostering accountability, ensuring inclusiveness and equity, and promoting AI that is responsive and sustainable as core ethical commitments for AI used in healthcare contexts (WHO, 2021).

Common Ethical Pitfalls

• Privacy lapses. Entering identifiable patient information into a general-purpose AI tool without an appropriate data processing agreement — one of the most common and most avoidable mistakes.
• Undetected bias. Assuming AI-generated content is neutral, when it can reflect skewed or unrepresentative training data — a documented concern across health-AI applications.
• Over-reliance. Letting AI convenience gradually erode independent clinical reasoning, especially under time pressure.
• Silent use. Using AI in ways that are never disclosed to the patient or client, even informally — undermining the transparency that ethical use depends on.
• Treating fluent output as validated. AI systems, including consumer chatbots, have been the subject of specific health advisories — including from the American Psychological Association — warning that generic, non-clinically-validated AI chatbots pose real risks when used for mental health support, particularly for vulnerable users, precisely because fluent and confident-sounding output can be mistaken for a validated clinical resource (APA, Health Advisory on the Use of Generative AI Chatbots, Wellness Apps, and Mental Health).

A Practical Pre-Use Ethics Checklist

Before using any AI tool in your clinical or therapeutic practice, ask:

• Have I removed or avoided entering identifiable patient information?
• Is this tool covered by my organization's data protection policy and, where applicable, a GDPR-compliant data processing agreement?
• Will I personally review and verify every clinically relevant claim in the output before it is used?
• Have I considered whether this output could reflect bias relevant to this specific patient's background, language, or presentation?
• Am I using AI to support documentation and materials — not to make a diagnostic, eligibility, or risk-related decision?
• Would I be comfortable explaining, in plain terms, how I used AI in this case if my patient or a regulator asked?

How Dr. Rania Kassir Approaches Ethical AI Training

Happy Brain Training's approach to ethical AI training centers on the same principle running through this article: AI should reduce administrative burden and support clinical work, never sit upstream of clinical judgment or patient safety. Every practical prompt or workflow taught in the training is paired with the ethical guardrails needed to use it responsibly, so that adoption of AI in practice happens with the same rigor clinicians already apply to any other clinical tool.

Regulatory Frameworks by Region

No single global law governs AI in healthcare. Instead, clinicians work within a patchwork of international guidance, regional regulation, and profession-specific rules — and the obligations that actually apply depend on where you practice, who you treat, and which professional order you answer to. The overview below is a starting map, not a compliance checklist: always confirm current requirements with your own regulator, professional order, and employer.

Global. The World Health Organization's 2021 guidance remains the most widely cited cross-border ethical framework for AI in health, and its six core principles — protecting autonomy, promoting safety and well-being, ensuring transparency, fostering accountability, ensuring inclusiveness and equity, and promoting responsive, sustainable AI — underpin most national and professional frameworks that followed it (WHO, 2021).

European Union. Two instruments matter most. The GDPR governs how patient and client data (including any data processed by AI tools) may be collected, stored, and used, with particularly strict conditions around health data as a "special category." The EU AI Act, entering into force in stages through 2026–2027, classifies most AI systems used in health and safety-critical contexts as "high-risk," imposing specific obligations on risk management, data governance, documentation, transparency, and human oversight for any AI system used in a clinical or health-adjacent workflow.

France. The Haute Autorité de Santé (HAS) publishes guidance specific to digital health and AI, and the CNIL (Commission Nationale de l'Informatique et des Libertés) is the data protection authority overseeing GDPR compliance for health data, including AI-assisted processing.

Belgium. The Institut National d'Assurance Maladie-Invalidité (INAMI/RIZIV) governs reimbursement and professional practice conditions, while the Centre fédéral d'expertise des soins de santé (KCE) publishes health-technology assessments relevant to digital and AI-based tools entering clinical use.

Luxembourg. The Caisse nationale de santé (CNS) and the Ministry of Health set the framework for health-data handling and reimbursed care; professional orders for each discipline (such as the Association Luxembourgeoise des Orthophonistes for speech-language therapy) issue discipline-specific guidance.

Switzerland. Switzerland is not an EU member state, so the GDPR and EU AI Act do not apply directly — Swiss data protection is governed by the revised Federal Act on Data Protection (FADP), overseen by the Federal Data Protection and Information Commissioner (FDPIC), which has published specific guidance on AI and data protection.

Canada. Health professions are regulated provincially. National and provincial professional orders — such as SAC-OAC nationally, the OOAQ in Québec and CASLPO in Ontario for speech-language pathology, or the OPQ in Québec and CPBAO in Ontario for psychology — set the practice standards that govern AI use within each discipline, alongside federal and provincial privacy legislation (PIPEDA and provincial equivalents).

United States. HIPAA governs the privacy and security of protected health information, including when AI tools process it, and any AI tool marketed as a medical device or diagnostic aid may fall under FDA oversight for AI/ML-enabled medical devices. Professional bodies — ASHA for speech-language pathology, APA for psychology, APTA for physical therapy — publish discipline-specific practice guidance on generative AI.

United Arab Emirates. The Dubai Health Authority (DHA) regulates healthcare delivery and data handling in Dubai, including requirements relevant to digital health and AI-assisted tools used in licensed clinical settings; other emirates have their own health regulators with comparable frameworks.

Explore our AI courses for beginners et online AI courses, naviguer Ressources gratuites — including our practical checklist below — learn more A propos de nousou Contactez-nous avec des questions.