Omnia
Omnia

Costa Rica just regulated AI in psychology, and it mirrors the APA and EU playbook

Patricio Espinoza

Patricio Espinoza

July 25, 2026

Costa Rica just regulated AI in psychology, and it mirrors the APA and EU playbook

Costa Rica just regulated AI in psychology, and it mirrors the APA and EU playbook

In 2026, Costa Rica's professional board of psychologists (CPPCR) issued two binding guideline documents: one on the ethical use of artificial intelligence, one on clinical record-keeping.

The short version: AI is allowed as a support tool, but AI-generated diagnoses and reports are prohibited, patient data must be fully anonymized before touching any AI system, and informed consent must explicitly disclose AI use.

If you practice outside Costa Rica, keep reading anyway, because this small country just wrote a preview of the regulation coming to your jurisdiction.

What exactly did the CPPCR publish?

Two documents, drafted by the board's oversight office (Fiscalía) and approved by its Board of Directors in 2026:

  • Guidelines for the management and custody of clinical records, minimum record content, informed consent, a 10-year retention duty, professional secrecy, telepsychology and information-security requirements.
  • Guidelines for the ethical use of artificial intelligence tools, 15 articles governing how psychologists may, and may not, use AI in professional practice.

The references section is the tell: the guidelines are explicitly built on the APA's 2023 ethical guidance for AI in health service psychology, the EU's trustworthy-AI framework, WHO's guidance on AI governance in health, and Costa Rica's data-protection law (Law 8968, the local GDPR analogue). This is not local improvisation, it is the international canon, made enforceable.

The core rules, in plain language

  • AI as support, never as substitute for professional judgment; the practitioner retains full ethical and legal responsibility.
  • No AI-generated diagnoses, diagnostic impressions or reports. Those require professional assessment.
  • No delegating professional decisions, interventions, referrals, evaluations, to AI systems.
  • No identifiable patient data in AI systems, directly or indirectly identifying; information must be fully anonymized first.
  • Exclusive-use AI accounts and tools with serious cybersecurity standards.
  • Informed consent must disclose AI use, and if the patient declines, the professional must not use AI in their case.
  • Mandatory validation: every AI output must be rigorously verified before entering professional practice, with sources cited.

Alongside it, the records guideline requires password-protected digital records, secure cloud storage, backups, access control, and a 10-year custody period, fully digital for telepsychology.

Why this matters beyond Costa Rica

Three reasons:

First, the convergence signal: when a national board operationalizes APA + EU + WHO principles into enforceable professional rules, it shows what the global baseline is crystallizing into, anonymization by default, human-in-the-loop by mandate, disclosure by consent.

Second, the generic-AI problem: under rules like these, pasting session notes into a general-purpose chatbot becomes multiply non-compliant (identifiable data, no exclusivity guarantees, no clinical safeguards). That pattern, common in practice today, is exactly what boards are moving against.

Third, the tooling implication: compliance shifts from being a matter of individual discipline to a matter of product design. Tools built for clinicians will need anonymization layers, hypothesis-framed outputs (never diagnoses), AI-disclosure documentation, and record security by architecture.

What compliant AI tooling looks like

Reading the CPPCR articles as a design specification, a compliant platform for psychologists: keeps the clinical record and the AI engine separated, so identified data never reaches the model; presents every output as a reference hypothesis for the professional's judgment, never as a diagnosis, prescription or decision; stamps AI-assisted documents with the corresponding disclosure; supports the informed-consent duty; and secures records with encryption, access control and retention aligned with professional rules. That is the design philosophy behind Omnia, our clinical-reasoning assistant for psychologists, built in alignment with these guidelines from the architecture up. You can see the article-by-article mapping on our compliance page.

This article is an independent analysis by Omnia of public documents issued by the Colegio de Profesionales en Psicología de Costa Rica (CPPCR). Omnia is not affiliated with, nor endorsed by, the CPPCR. Always consult the official guidelines at psicologiacr.com.

Patricio Espinoza

About Patricio Espinoza

Psicoterapeuta integrativo fundador de Omnia