Page 9 - Shaping ethics, regulation and standardization in AI for health

Executive summary

The Focus Group on Artificial Intelligence for Health (FG-AI4H) was operated from 2018 to
2023 as a partnership of the International Telecommunication Union (ITU) and the World Health
Organization (WHO) to develop a standardized assessment framework for the evaluation of
AI-based methods for health, diagnosis, triage or treatment decisions in line with the United
Nations’ Sustainable Development Goals. Participation in the FG-AI4H was free and open to all.

The FG scope and general process were described in a Commentary in The Lancet [6] and in a
white paper [4]. An onboarding document was prepared to help newcomers participate in the
FG-AI4H work. Its detailed terms of reference are found in https:// itu .int/ en/ ITU -T/ focusgroups/
ai4h/ Documents/ FG -AI4H -ToR .pdf.

After five years in operation, the FG-AI4H reached its sunset with the following results:

– A dynamic, inclusive, interdisciplinary/cross-sector community of international experts that
met regularly online as well as in 19 meetings over the FG-AI4H lifetime.
– A set of 36 completed deliverables [1] that were submitted to ITU-T Study Group 16 (now
Study Group 21) to facilitate the development of technical standards on AI for health.
Various deliverables were also published by WHO.
– Uniquely valued guidance focused on the use of AI in health, including aspects such as
ethics, regulatory guidance, data quality and clinical evaluation (DEL1 to DEL7 series).
– A deliverable with guidance on the use of AI and other digital technologies within the
context of the COVID-19 pandemic, which may be useful when facing future health
emergencies.
– Exploration of a wide range of health use cases with application of the guidance developed
by the FG-AI4H (DEL10 series).
– The Open Code Initiative (OCI), a cloud software platform (https:// github .com/ fg -ai4h)
that provides a proof-of-concept of the application of the various FG-AI4H horizontal
deliverables for specific health use cases studies under the 24 Topic Groups of the FG-
AI4H.

It took some time to build this international community of experts as well as to properly scope
the issues to be tackled. Special considerations were given to AI in health standardization in Low-
and-Middle Income Countries (LMICs). The FG-AI4H deliverables provide a solid starting point
of a benchmarking framework for equitable and inclusive access to AI for health technology,
but the work is far from complete as we see fast evolution on AI field is in rapid evolution. There
is value in continuing the current momentum to further the conditions for safe and ethical
adoption of AI for health worldwide, also considering the fast evolution of AI technologies and
the increasing interest of governments, industry and academia in applying AI to health; for
example, large language models became a popular topic when the FG-AI4H was already close
to its closing and were not addressed at all by the group.

ix