Policy 50 — Responsible Use of Artificial Intelligence

← All 75 policies · Full framework PDF · Forms · Print this policy · Governing text: English · Reviewed 6 October 2026

PHIG Policy Framework v1.2 · Policy 50 · Adopted 1 September 2026 · References: EU Artificial Intelligence Act (Regulation (EU) 2024/1689) risk-based principles; UNESCO Recommendation on the Ethics of AI; WHO guidance on ethics and governance of AI for health (2021, 2024); COPE/ICMJE guidance on AI in publishing; US OMB M-24-10 principles

1. Purpose and scope

PHIG uses AI systems to find, draft, translate and structure information — the GILS Desk, the Labs Desk, the GMJ selection, the Sheni Labs verdict process, course production. This policy ensures that such use is transparent, accountable, lawful and under human control. Applies to the Board, the Director, staff, consultants, volunteers, experts, partners and suppliers of PHIG and all platforms of its network.

2. Policy

1. Human accountability: every output published, submitted or acted on is reviewed and approved by an identified person who is responsible for it; AI is never an author, assessor, editor or decision-maker of record.
2. Transparency: automated systems are described publicly (sources, rules, exclusions, review); content substantially produced with AI assistance states so; AI tools used in research are disclosed in methods.
3. Sources only: automated systems draw on published, identified sources; they do not invent facts, quotes or references; every figure remains traceable.
4. Data protection: personal data and confidential information are not entered into AI tools unless the tool is contracted as a processor under the Data Protection Policy; health data of individuals is never processed by general-purpose AI tools.
5. Risk-based control: uses that affect individuals’ rights or safety (assessment of a person, access to a service, health advice) are classified as higher-risk and require documented human review, bias testing and the possibility of appeal; prohibited uses (manipulation, social scoring, covert profiling) are not undertaken.
6. Quality: AI outputs are checked for accuracy, bias and currency; errors found are corrected and logged; models and prompts of PHIG’s automated desks are versioned.
7. Excluded topics and editorial rules of the network apply to AI systems as to people.
8. Procurement of AI tools follows the Procurement and IT Security policies; vendor terms on data use are reviewed.

3. Procedures

1. Register of AI systems in use with owner, purpose, risk level and review; annual review of each automated desk.

Responsibilities

Director; platform editors; Board for higher-risk uses.

Review

Annually, given the pace of regulation.

Part of the PHIG Policy Framework. Breaches and concerns may be reported under the Whistleblowing Procedure to info@accreditation.ge (subject “Confidential — integrity”).