Page 70 - The Annual AI Governance Report 2025 Steering the Future of AI

The Annual AI Governance Report 2025: Steering the Future of AI

2.7 Sectoral Focus and Broad Collaboration

AI is not a single technology but a family of systems with sector-specific impacts. Governance
must therefore be sensitive to context. In healthcare, safety and explainability are paramount.
In education, equity of access is key. In agriculture, resilience and food security are priorities. Pillars Chapter 2: Ten

Examples shared during the Dialogue illustrated how sectoral governance can work in practice:
AI-assisted dementia screening projects, registries of AI use in public administration, and
agricultural platforms monitoring soil and water conditions. Participants stressed that sectoral
governance should not fragment into silos, but instead serve as testbeds for principles that can
then be scaled globally.

Quote:

• “By today, Estonia has implemented approximately 200 AI applications across
government institutions – in fields as diverse as education, healthcare, justice,
transport, the environment, and culture.” (H.E. Mr. Alar Karis, President of the
Republic of Estonia)

Dive deeper in the Whitepaper “Themes and Trends in AI Governance”:

• 3.3 Regional AI partnerships
• Annex – Examples of multilateral initiatives [a list of some 40 initiatives]
• Annex – Examples of national initiatives [a list of some 20 initiatives]

2.8 Standards and Safety Tools

Standards were identified as one of the pressing needs for global governance. Without common
benchmarks for testing, evaluating, and registering AI systems, efforts risk fragmentation. There
was strong support for international cooperation on standards, especially for safety testing of
frontier models.
It was also said that, historically, developing countries have been largely excluded from
standard-setting processes, leading to barriers for smaller market players. Multistakeholder
and multidisciplinary approaches to testing that incorporate diverse perspectives and local
realities, especially regarding how technologies fail in different parts of the world, will be useful.

The need for greater harmonization in standardization efforts to reduce fragmentation and
to maintain flexibility to keep pace with rapid AI advancements was pointed out. One needs to
double down on benchmarking efforts to test AI's robustness, safety, and fairness, especially
for industry-specific applications (healthcare, education, defense, etc.) and emerging agentic
AI solutions with greater autonomy.

Risk management systems for frontier AI need to be standardized, emphasized Chris Meserole
(Executive Director, Frontier Model Forum), and called for global coordination to define what

61