Multiple international AI governance initiatives are being put in place, including UNESCO’s Global AI Ethics and Governance Observatory,27 the Readiness Assessment Methodology (RAM) tool,28 the Hiroshima Principles29 and the OECD’s ethical AI governance framework.30 While these efforts lay a strong foundation, there remains a critical need for further action and broader agreement on global ethics frameworks for AI. Responsible use guardrails Responsible use guardrails promote the ethical and responsible management and use of AI across various sectors, helping to prevent harmful applications while maintaining public trust and accountability. The recently published USAID AI Action Plan report notes the need for significant stakeholder engagement, with governments providing strategic vision, and academia addressing complex challenges with civil society. Establishing responsible AI practices requires a thoughtful approach to ethical standards, comprehensive transparency initiatives and a continuous dedication to societal improvement in various technological domains. The challenge encompasses not only the creation of these standards but also the cultivation of public trust and the maintenance of accountability amid the swift progression of AI technologies. Self-governance tools have been widely adopted by the largest developers of AI models, such as Microsoft’s Responsible AI Principles,31 Google’s AI Principles32 and Salesforce’s Office of Ethical and Humane Use.33 The promotion of self-governance processes should be encouraged within the small and medium-sized technology business community. However, self-regulation presents a host of challenges such as limited oversight and accountability. Self- regulation alone can sometimes be insufficient and necessitates a degree of governmental intervention to ensure consistency in the implementation of responsible and ethical AI principles. Safety and security standards AI poses risks that are both known and still emerging, particularly as researchers progress towards advanced AI development, such as artificial general intelligence (AGI). To mitigate these AI safety and security risks, it is crucial to first establish clear policy “red lines” and safety guardrails. The EU Artificial Intelligence Act34 categorizes AI applications into risk levels, setting requirements for high-risk areas like critical infrastructure while promoting innovation in low-risk sectors. This approach defines “red line” areas where AI poses unacceptable risks. In another example, the collaboration between the US and UK, through their AI Safety Institutes,35 focuses on developing shared frameworks for testing advanced AI models, emphasizing international collaboration. The NIST AI risk framework36 is an example of a voluntary structure for managing AI risks, emphasizing trustworthiness and alignment with international standards but still lacking in enforceability and global consensus. A global framework and international body for AI could set boundaries on high-risk technologies, such as autonomous weapons and mass surveillance systems. The recent collaboration between the UK, US and Canada on AI in the nuclear sector37 highlights the importance of international cooperation, emphasizing risk management and balancing human oversight with AI autonomy. AI regulations The rapidly changing regulatory landscape for AI presents significant challenges for industries delivering technological advancements at regional or global scale. Companies must adapt to complex and evolving AI-specific regulations across various jurisdictions while ensuring compliance with data protection laws and industry standards. Regulatory approaches vary widely, from hands-off to hands- on, and can differ even within regions. A hands-off approach to regulation minimizes government intervention, allowing for rapid innovation and market-driven growth by reducing barriers to entry. While this creates an environment conducive to experimentation, it has led to public concerns over privacy violations and the misuse of technologies like facial recognition. The alternative is a hands-on approach that promotes government intervention with clear guidelines and accountability mechanisms. The EU Artificial Intelligence Act is one example, setting regulated requirements for high-risk AI applications, aiming to protect public interests while encouraging innovation through structured oversight. Narrowly targeted regulation by governments can also be a valuable policy lever and can proactively prevent emerging AI risks while supporting innovation. The World Economic Forum’s Governance in the Age of Generative AI report38 suggests that governments should enhance existing regulations, clarify authorities and assign responsibilities to adapt to AI’s regulatory challenges. This includes addressing privacy, consumer protection, product liability and competition issues. Establishing responsible AI practices requires a thoughtful approach to ethical standards, comprehensive transparency initiatives and a continuous dedication to societal improvement. Blueprint for Intelligent Economies 14

A Blueprint for Intelligent Economies 2025