Implementome logo

EmpAIth Assurance Framework (EAF™) (Clinician-Led Governance Model for Behavioral-Health Artificial Intelligence)

Origin language: English
Cover image
Target population
Adults (18+) , +3
Beneficiaries1 - 50
EvidenceBoth
Start date2025/07
End dateOngoing
StagePilot project/testing/trials
Country
United States of America
Coverage
International
Nb of implementation sites
1
Lead organization
Implementation partners
N/A

Health Focus Area
Mental health
Primary Health Care
Health promotion
Others
Services
Telehealth
Emergency preparedness and response
Enabling technologies
Other
Artificial Intelligence
Standards
XDS - Cross-Enterprise Document Sharing
ATNA - Audit Trail and Node Authentication
SNOMED
PII
CDA - Clinical Document Architecture
HL7 v3
ISO 3166
ICD-11
Tools
N/A
Funding sources
Private funds
Business model
Others
Usage fee / pay per use
Data monetization
Funders
N/A
Summary
The EmpAIth Assurance Framework (EAF™) is an IRB-approved governance model that sets measurable clinical, ethical, and regulatory standards for artificial intelligence systems in behavioral health. Developed under the oversight of WCG IRB (Protocol CC-2025-001), EAF™ offers a structured rubric and auditing methodology to assess AI tools used in mental health and psychosocial support. It aligns with professional ethics and data protection requirements, incorporating guidelines from the NIST AI RMF (2023) and EU AI Act (2024) across ten auditable governance domains, such as Clinical Validity, Bias & Fairness, Crisis Detection & Response, and Privacy & Confidentiality. By introducing licensed clinician oversight, EAF™ promotes trustworthy, human-in-the-loop design, enhancing user safety, cultural responsiveness, and regulatory readiness. It is suitable for adoption by health ministries, research institutions, and developers seeking reliable, clinician-validated digital mental health technologies. The framework aims to scale a clinician-led, IRB-approved AI governance model that emphasizes safety, transparency, and accountability in behavioral health technology, while training clinical reviewers globally and aligning policies with WHO AI ethics and digital health standards. Target populations include health professionals, adults, adolescents, and children.
Keywords
Mental health
Artificial intelligence
Newborn health
Digital health initiatives
Digital health solutions
Publications
EmpAIth Assurance Framework (EAF™): A Clinician-Led Governance Model for AI in Behavioral Health
Insights - Lessons learnt
Challenges
• Lack of existing clinical-grade AI audit frameworks: At the time of implementation, few standardized or validated governance tools existed for behavioral health AI, creating challenges in benchmarking and aligning with global regulatory expectations. • Technical variability across AI systems: Behavioral health AI tools vary significantly in architecture, data sources, and model maturity, which required adaptable scoring logic and flexible evaluation pathways. • Limited stakeholder familiarity with AI governance: Many clinicians, administrators, and decision-makers had limited exposure to structured AI audit processes, requiring additional explanation, training, and capacity-building. • Emerging regulatory landscape: Rapid changes across NIST, the EU AI Act, and WHO guidance required continuous updates to keep the framework aligned with the most current standards. • Sensitivity of mental-health-related data: Psychological and behavioral data present heightened ethical risks, necessitating rigorous privacy, safety, and confidentiality safeguards during validation.
Recommendations
Positive Influencing Factors: • IRB oversight and structured methodology (WCG IRB CC-2025-001) increased trust, transparency, and acceptance across early users. • Clinician-led development strengthened user confidence, ensuring the criteria were aligned with real-world care and safety needs. • Transparent documentation and open science practices (Zenodo DOIs) accelerated external review and citation, supporting scalability and replicability. • Alignment with international standards (NIST AI RMF, EU AI Act, WHO guiding principles) facilitated rapid integration across healthcare ecosystems. Recommendations: • Invest in clinician and stakeholder capacity-building to ensure safe, informed adoption of AI systems in mental and behavioral health. • Maintain a continuous-update governance process to remain aligned with evolving international regulations and safety expectations. • Apply the EAF™ as a reusable audit tool for hospitals, digital health programs, and global health organizations to strengthen quality assurance. • Promote open science practices by pairing every audit with transparent documentation, reproducible methodology, and publicly accessible supplementary materials. • Prioritize high-risk AI systems first (behavioral health, crisis detection, vulnerable populations) to maximize immediate safety impact.
Recommendations
Projects using similar services
Projects focusing on similar topics
Projects with similar insights
Any additional questions?
iDHA (Project ID)
ykjoan
Project links
https://empaith.com
Origin of information
Publication report
Data source link
https://doi.org/10.5281/zenodo.17489429
Added to the platform on
2025-11-13
Project editors
Last update by
Alexis Baudin on 2025-11-19
Number of views
55
WHO classifications
WHO LogoGeneva Digital Health Hub Logo