Page 55 - AI Standards for Global Impact: From Governance to Action

AI Standards for Global Impact: From Governance to Action

GI-AI4H is a collaborative project to harness the power of AI for global health improvement.
The initiative aims to develop international standards, norms, and policies to guide the ethical
and responsible use of AI in health, promote knowledge and data sharing, foster collaboration,
and build a global community of AI for Health experts. Building upon the framework with 36 Part 2: Thematic AI
deliverables from the FG-AI4H, the Initiative also seeks to establish sustainable models for
implementing AI programmes at the country level, making AI solutions accessible and impactful
across diverse health systems, leveraging three pillars:

1) Enable:
o Norms, guidance, standards (including all benchmarking framework deliverables)
o Governance (ethics, regulations)
o Surveillance, research and evidence

2) Facilitate:
o Knowledge sharing between countries
o Pool funding
o Cooperation between all stakeholders
3) Implement:

o Scale programme in countries (phase 1 to target 12 to 18 countries)
o Build capacity building for AI for health programmes
o Build sustainability models

8�2 Strategy and Frameworks for AI in health

On the policy and governance challenges and opportunities associated with the adoption of
strategic frameworks for AI in health, topics discussed ranged from regulatory harmonization,
ethical implementation, and equitable access to how GI-AI4H can facilitate global collaboration.

The following challenges in implementing AI in health were highlighted by speakers during
this session:

a) AI for health must be customized to local needs and infrastructure. Meanwhile, it should
evolve to ensure transparency, accountability, and ethical oversight.
b) The importance of localization and sovereignty of LLMs based on experience in developing
contextualized, trustworthy, and transparent LLMs.
c) The EU’s efforts to tackle medical data management and its proposed principle-based
approach enabling agile adaptation to technological changes.
d) There is a need to build communities and champion networks to guide AI deployment.
e) AI should be viewed as an enabler, not a replacement, in health.
f) Global collaboration is vital to share best practices and avoid duplication.
g) Investments are needed in research, infrastructure, regulation, and human capacity.
h) Human-centred validation of AI models is essential for trust and adoption.
i) Standards are needed for the validation of models and must support transparency,
reproducibility, and contextual validation.

The following action areas were identified for strategy and frameworks for AI in health:

1) Develop agile, standards-based regulatory templates to help countries adapt AI
frameworks to local health needs while ensuring global alignment.

43