From Principles to Practice: The Global Index on Responsible AI and the State of AI Governance

AI is advancing at a pace few governments anticipated. Once confined to research laboratories and niche industries, AI systems are now embedded in public administration, healthcare, education, finance, policing, employment, and the delivery of essential public services. Governments across the world have responded by publishing national AI strategies, ethical principles, and governance frameworks intended to promote responsible innovation. Yet an important question remains: are these frameworks translating into meaningful governance?

The Global Index on Responsible AI (GIRAI) 2026 provides one of the most comprehensive attempts to answer that question. Rather than measuring technological capability or AI investment, the Index evaluates whether countries have established the legal, institutional, and policy foundations necessary to ensure that AI is developed and deployed responsibly. Covering 135 countries, the report assesses governance across multiple dimensions, including AI policy, public administration, human rights, labour, environmental sustainability, and access to remedy.

The report reaches a sobering conclusion. While AI investment and adoption continue to accelerate worldwide, governance capacity has failed to keep pace. Global corporate investment in AI nearly tripled between 2023 and 2025, reaching approximately US$581.7 billion, while more than half of the world's population now uses generative AI in some form. Despite this rapid transformation, the average GIRAI score across all assessed countries remains only about 35 out of 100, demonstrating a significant gap between technological advancement and public governance.

The Index reveals that governments are increasingly recognising the need for responsible AI governance, but recognition has not been matched by implementation. A total of 128 of the 135 countries assessed have adopted at least one responsible AI framework or policy commitment. Across all indicators assessed, however, evidence of actual implementation exists in only 55 per cent of cases globally, falling to 45 per cent in the Global South. Many governments have adopted strategies without establishing the independent oversight bodies, monitoring systems, enforcement mechanisms, or avenues for public accountability necessary to make those commitments meaningful in practice.

This implementation gap is particularly evident in the public sector. The GIRAI identifies AI use in public service as the weakest-performing dimension across the entire Index. Only a small proportion of countries have established governance frameworks requiring transparent procurement of AI systems or public disclosure of government algorithms. This is significant because governments increasingly rely on AI when making decisions that affect access to healthcare, education, welfare benefits, immigration, policing, and other public services. When the state deploys AI without adequate safeguards, the consequences extend beyond administrative efficiency to fundamental rights and democratic accountability.

Several jurisdictions nevertheless demonstrate what meaningful implementation looks like. Brazil stands out as one of the few Global South countries that combines policy commitments with concrete government action. Its Artificial Intelligence Strategy integrates ethical considerations into public procurement, while complementary initiatives support responsible acquisition of AI systems by public institutions. Brazil has also established government programmes examining AI's impact on employment, adopted sustainability measures for AI infrastructure, and issued guidance governing the use of generative AI within public administration. Rather than treating responsible AI as an abstract policy aspiration, Brazil has embedded governance requirements throughout the AI lifecycle.

Chile has similarly developed detailed procurement rules requiring transparency, algorithmic auditing, bias mitigation, and data protection throughout the acquisition of AI systems. Australia has adopted a National Framework for the Assurance of Artificial Intelligence in Government that requires accountability mechanisms, performance testing, government capability assessments, and knowledge-transfer arrangements designed to prevent excessive dependence on technology vendors. These examples illustrate that responsible AI governance extends well beyond publishing ethical principles; it requires institutions capable of supervising AI systems before, during, and after deployment.

Europe provides another important point of comparison. Much of the increase in enforceable responsible AI protections identified by the GIRAI is attributable to the European Union's AI Act, which establishes legally binding obligations for developers and deployers of high-risk AI systems. The report contrasts this with many Global South jurisdictions, where responsible AI governance continues to rely primarily on soft-law instruments such as strategies, guidelines, and policy statements. While these documents play an important role in setting national priorities, they generally lack the enforceability necessary to guarantee individual rights or hold public and private actors accountable for harmful AI practices.

The African experience reflects both encouraging progress and persistent challenges. Several countries have adopted national AI strategies, data governance frameworks, or digital transformation policies aimed at positioning AI as a driver of economic development. Kenya, Rwanda, South Africa, Nigeria, and Egypt have all taken significant steps toward developing AI ecosystems, while continental initiatives led by the African Union seek to promote harmonised governance principles across member states. Yet, as the GIRAI demonstrates, the existence of policy frameworks does not necessarily translate into institutional capacity. The challenge for many African governments is no longer whether AI should be governed, but how governance commitments can be operationalised through legislation, regulators, public institutions, technical expertise, and sustainable funding.

Kenya illustrates both the opportunities and the limitations identified by the Index. In recent years, the country has introduced important legal and policy instruments, including the Data Protection Act, the National AI Strategy, and the Draft National Data Governance Policy. These initiatives demonstrate growing recognition that AI requires a coherent governance framework capable of balancing innovation with the protection of fundamental rights. However, as the GIRAI suggests, future success will depend less on producing additional policy documents and more on ensuring effective implementation through independent oversight, transparent procurement practices, regulatory enforcement, and meaningful public participation.

Ultimately, the GIRAI does not measure which countries possess the most advanced AI technologies. Instead, it measures whether governments are building the institutions necessary to govern AI responsibly in the public interest. Its findings reveal a global governance paradox: AI adoption is accelerating, while public institutions responsible for overseeing AI continue to lag. Closing this implementation gap will require governments to move beyond aspirational strategies towards enforceable legal frameworks, adequately resourced regulatory bodies, and transparent governance systems capable of earning public trust.

The Global Index on Responsible AI, therefore, offers an important reminder that responsible AI cannot be achieved through policy declarations alone. As artificial intelligence becomes increasingly embedded in decisions affecting people's rights, livelihoods, and access to public services, the effectiveness of governance will ultimately be judged not by the number of strategies governments publish, but by the strength of the institutions they build to ensure those strategies work in practice.