Transparency and explainability, accountability, fairness and inclusivity, privacy, trustworthiness, and safety are well-established AI principles that have reached near-global consensus, since as early as 2021, when 193 countries supported the UNESCO Recommendation on the Ethics of Artificial Intelligence. 

But what have countries done to implement these principles in policy and action? 

The Center for AI and Digital Policy (CAIDP) tracks AI policy and practices across 90 countries in the annual Artificial Intelligence and Democratic Values Index (CAIDP AI Index). CAIDP’s mission is to promote a better society, more fair, more just, and more accountable world—a world where technology promotes broad social inclusion based on fundamental rights, democratic institutions, and the rule of law. 

We have a global footprint across 130 countries with a network of more than 2,000 AI policy experts, including alumni and members of our Global Academic Network (GAN) since 2021.

The Index combines qualitative country reports with metrics grounded in human rights, democracy, and the rule of law, offering a comparative view of how governance is evolving in practice. The 2026 CAIDP AI Index, released in April, shows measurable progress toward aligning AI governance with democratic values and fundamental rights. 

While the United States declined by 1.5 points, most countries moved into higher tiers, pointing to steady global progress in AI governance. Concrete policy actions—from new comprehensive legislation in countries such as Korea, Italy, and Peru, to advancing AI literacy and capacity-building programs for both the public and civil servants—show how governments are beginning to move from principles to practice.

The direction is clear, even if progress remains uneven.

Methodology

Operationalized as questions, the CAIDP AI Index metrics synthesize the extent of concrete policy actions in AI governance that reflect AI policy frameworks and democratic decision-making.  

The 12 metrics provide the basis for longitudinal (across time) and cross-sectional (vs other countries, within or between regions) comparative evaluation.

The 12 metrics provide the basis for longitudinal (across time) and cross-sectional (vs other countries, within or between regions) comparative evaluation.  We base the scoring on a rigorous review of publicly available, authoritative information. 

Below, we present quantitative findings and reflect on their implications for public sector workers.

Where Countries Stand on AI Governance Metrics

In 2025, the CAIDP AI Index rankings signaled widespread adoption of AI governance aligned to democratic values and the protection of fundamental rights. 

A majority of countries (61%) have endorsed the OECD AI principles (Q1), and 93% have at least partially implemented them (Q2). Similarly, 93% have at least partially implemented the UNESCO Recommendation on AI Ethics.

Only five countries—Brazil, Chile, the Netherlands, Peru, and Saudi Arabia—met both requirements of completing the UNESCO readiness assessment and taking additional steps toward ethical AI deployment.

With the 2026 edition, CAIDP now requires completion of the UNESCO readiness assessment to earn full implementation points. Only five countries—Brazil, Chile, the Netherlands, Peru, and Saudi Arabia—met both requirements of completing the assessment and taking additional steps toward ethical AI deployment.

As indicators of robust governance mechanisms, 91% of countries implement some form of meaningful public participation in the development of national AI policies (Q5). 98% make AI policies and practices available to the public in part or in full (Q6), and 77% of countries have some form of supervisory authority or mechanism for independent AI oversight (Q7). 

An increasing number of countries (94%) explicitly include fairness, accountability, transparency, rule of law, and/or fundamental rights in national AI policies (Q8); 50% include all five characteristics. An auspicious 80% have established a right to algorithmic transparency by law, though only 49% include a right to human review or the ability to contest adverse decisions, as reflected in frameworks like the General Data Protection Regulation and the Convention 108+ (Q9).

Endorsement of the Council of Europe (CoE) international AI treaty, the first binding convention on AI, human rights, democratic values, and the rule of law (Q10), and sponsorship of AI-related resolutions at the Global Privacy Assembly level (Q12) remain the least adopted measures.

Shifts since 2024

A generalized shift to higher tiers suggests continued progress in rights-based AI governance and democratic decision-making.  Canada and Japan led the rankings in the 2026 CAIDP AI Index, with the Netherlands, Norway, Switzerland, and the United Kingdom rounding out Tier I. 

All countries in the top tier support the international AI treaty developed in the CoE. South Korea, which ranked in Tier I in the 2025 edition, moved to Tier II after losing points for UNESCO implementation due to the country not completing the readiness assessment.

Tier II expanded, with five countries moving up from Tier III (excluding Romania, newly added in 2026).  Of the other countries added for 2026, Greece and Slovakia scored in Tier III; Benin, Jordan, and Oman in Tier IV; and Ecuador, Mozambique, Namibia, and Tanzania in Tier V.

Excluding new entrants, Tier V decreased significantly, while Tiers IV and III each saw a net decrease of 1 country. 

Faltering democratic institutions and a lack of protection for fundamental civil and political rights are common among countries in the lower tiers. Of the countries in Tier V, only Namibia and Trinidad and Tobago are considered “free” in the Freedom House rankings. Ecuador, Mozambique, and Thailand receive partial points on that measure. Countries in Tiers IV and V tend to have less robust mechanisms for meaningful public participation in AI policymaking. 

Notable Individual Falls/Rises

The U.S. abdication of leadership in global AI governance and wavering on democratic values accounted for the country’s fall to the bottom of Tier III after promising advances in 2024. The centralization of power under the executive branch, attempts to preempt state legislation, and responsiveness to corporate input over civil society contributed to a lower score on public participation (Q5).

The withdrawal from US commitments to international organizations, specifically UNESCO, also led to a decrease (Q11). The Trump Administration’s rescission  of earlier executive orders and policies and lack of attention to fairness, transparency, accountability, fundamental rights, and the rule of law in new policies led to a decrease in national AI policies (Q8) 

The Philippines, the only other country to fall by 1.5 points, improved in some metrics and fell in others. The overall score declined partly due to a prior misclassification regarding endorsement of OECD AI Principles.

More countries saw notable improvements. A national AI strategy and related policies addressing fairness, transparency, accountability, rights, and rule of law and engagement with the UNESCO Recommendation boosted Kazakhstan by 2.5 points, pushing the country to the top of Tier V.  New legislation and stronger implementation also led Ukraine, Senegal, and Taiwan to improve by 2 points each.

Implications for Public Sector Workers

CAIDP rankings underscore the role of public-sector workers in translating democratic values and governance frameworks into practice. Where oversight bodies exist, civil servants are responsible for monitoring systems and enforcing accountability.

The effectiveness of oversight agencies or offices depends on sustained capacity-building and training.

While 28% of countries have independent oversight bodies established by law, and 49% have some form of oversight mechanism, the effectiveness of these agencies or offices depends on sustained capacity-building and training. Oversight bodies and supervisory authorities must be familiar with the impact of AI systems and understand their technical functions.

CAIDP began tracking this capacity-building in the public sector and AI literacy for the general public in 2026. CAIDP also supports training efforts through partnerships with UNESCO and across the European Union because of our recognition of how critical the contribution of these workers are to progress. The CAIDP AI Policy Clinics offer another resource for practical AI policy training, with at least 20% of participants from local or federal governments. 

Beyond oversight, public sector employees and supervisory authorities shape how governance functions in practice, managing participation processes, maintaining transparency, and reviewing automated decisions. They are often the final decision-makers, for example, in determining eligibility for public benefits when automated systems are involved.

With nearly 1700 pages of detail on AI policies and practices across 90 countries and from leading institutions, the CAIDP AI Index aims to serve those making policy as well as those carrying it out. 

What the Index Shows When Taken Together

Governments have largely agreed on AI governance principles. Most countries now reference fairness, accountability, and transparency in policy; many have endorsed international frameworks and published strategies. In most places, the baseline has shifted from whether AI should be governed to how. 

The movement across tiers in the CAIDP AI Index shows gradual progress in implementing these principles. What varies is execution. Governments are still uneven in building the systems that make commitments to these principles real. The gap between partial and full implementation—visible in metrics like UNESCO readiness assessments or rights to contest automated decisions—marks the difference between symbolic alignment and operational governance. 

Countries in lower tiers tend to struggle with broader institutional constraints: weaker democratic systems, limited civil society participation, and fewer accountability mechanisms. AI governance often reflects these underlying conditions. The framing of an AI race and the related myth that regulation hinders innovation and development further obstruct progress.

Governance does not happen at the level of principles or rankings. It happens in agencies, through processes, and in decisions made by civil servants and supervisory authorities. 

For the public sector, the implication is practical. Governance does not happen at the level of principles or rankings. It happens in agencies, through processes, and in decisions made by civil servants and supervisory authorities. 

The CAIDP AI Index points to where those governance systems exist, where they are missing, and where the next layer of work is needed.

We encourage public sector workers worldwide to review their local legal and regulatory commitments and take steps to implement them. Encourage robust oversight. Request the capacity-building and tools you need to succeed in your work.