Page 10 - The Annual AI Governance Report 2025 Steering the Future of AI

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

Part I – White Paper: Themes

and Trends in AI Governance 1 Part I – White Paper

Executive Summary

An area of emerging focus in AI governance over the past year have been AI Agents. The rise
of AI agents – systems that can autonomously perform multi-step tasks, interact with software
environments, and make decisions with minimal human input – has introduced new governance
concerns, including traceability, agent coordination, and security vulnerabilities. As these agents
begin to be integrated into consumer tools and enterprise workflows, questions of oversight
and liability have become more urgent.

AI is also transforming science and innovation itself. AlphaFold - recognized with a Nobel Prize
in 2024 – continues to evolve and demonstrates the transformative role of AI in accelerating
protein structure discovery and fundamental research. Recent iterations of AlphaFold and other
emerging “AI scientist” systems are driving progress in molecular design, materials science, and
bioengineering. Meanwhile, autonomous experimentation platforms are beginning to integrate
robotics with language models to iteratively design, test, and refine scientific hypotheses, offering
an early glimpse into new research workflows shaped by AI-driven planning and reasoning.

On the socioeconomic front, AI deployment is reshaping labor markets not just through
automation, but by reorganizing the value of human work. While early concerns focused on
displacement in gig sectors – such as translation, design, and transcription – recent studies
suggest a more complex picture. More recent research shows that generative AI is augmenting
mid-skill roles, particularly in customer service and professional writing, leading to increased
productivity and wage compression rather than pure substitution. However, the displacement
effect remains concentrated in routine, low-mobility roles, with freelancers on digital platforms
reporting declining demand in certain categories. In parallel, large employers are experimenting
with hybrid workflows that pair AI systems with human oversight, suggesting a shift from full
substitution to "recomposition" of labor. Public sentiment increasingly supports automation
that removes repetitive or low-value tasks, but resists its use in sensitive or relational domains
(e.g., education, healthcare). To manage this transition, governments are piloting new forms of
digital public infrastructure—from India’s AI Skilling Stack to targeted worker transition funds in
the EU—aimed at embedding human-centered design and retraining pathways into the future
of work.

Over the past year, AI infrastructure, with compute resources being a critical pillar, has become
a central arena for strategic competition, industrial policy, and digital sovereignty. According
to an analysis of researchers at the University of Oxford, the United States, China and the
European Union account for over half of the world’s most powerful data centers. American and
Chinese companies operate more than 90 percent of the data centers used globally by other
organizations for AI work. Africa and South America have almost no computing hubs. India has
at least five; Japan at least four. More than 150 countries have none at all.

1 The white paper was sent to the participants of the AI for Good Dialogue luncheon (10 July 2025) prior to
the luncheon.

1