Artificial intelligence (AI) is spreading across the world faster than governments can regulate it and leaving significant gaps in public oversight, accountability and citizen protections, according to the 2026 Global Index on Responsible AI (GIRAI).
The report, released by the Global Center on AI Governance, found that while 128 of the 135 countries assessed have made some commitment to responsible AI governance, the average global score remains just 35 out of 100, revealing that policy implementation is lagging far behind the rapid deployment of AI systems.
The findings come as generative AI adoption reaches 53 percent of the global population and corporate investment in AI nearly triples from about $200 billion in 2023 to $581.7 billion in 2025.
AI systems are embedded in public services, education, healthcare, policing, finance and workplaces, the report noted.
Despite the proliferation of AI strategies and ethical frameworks, researchers found that evidence of implementation exists in only 55 percent of cases where responsible AI frameworks are active globally. The figure falls further to 45 percent across Global South countries.
“Responsible AI governance evolution has shown the limits of the framework-building phase,” the report said, arguing that governments must now shift from policy commitments to enforceable protections, transparent AI use and mechanisms that allow citizens to challenge AI-driven decisions.
The index noted that governments have been quicker to introduce AI literacy programmes than institutions capable of enforcing responsible AI.
Among 104 countries with active frameworks, researchers found that training courses and workshops were the most common implementation activities, while only 28 countries had independent oversight bodies, 22 had monitoring systems and just 16 had multistakeholder advisory processes.
One of the weakest areas identified was the use of AI in government itself as only 18 percent of countries require public disclosure of government algorithmic systems, while just 26 percent have frameworks governing the public procurement of AI technologies.
Researchers argued that governments are demanding transparency from technology companies without applying the same standards to their own AI deployments.
The report also found out that AI governance remains uneven across regions. Although Global South countries have significantly expanded responsible AI policies since the first edition of the index, most rely on voluntary or non-binding frameworks.
Framework coverage across Global South countries increased by 83 percent between editions, compared with a 35 percent increase in developed economies. However, 78 percent of responsible AI frameworks in the Global South remain non-binding, compared with 42 percent in the Global North.
The index further highlighted several governance blind spots. While AI literacy has become one of the strongest-performing indicators, labour rights, environmental protections, children’s rights and transparency in government AI use remain poorly addressed.
Only 39 countries have labour protection frameworks related to AI, while just 27 percent of countries have frameworks covering AI’s environmental impact.
The researchers also found credible evidence that governments in 35 of the 135 countries assessed have deployed AI systems posing unacceptable risks to human rights, including uses linked to surveillance, discrimination and other high-risk applications.
The report noted that the next phase of AI governance must prioritise enforceable regulation, stronger oversight institutions, transparency in public-sector AI, protection for workers and vulnerable groups, and greater cooperation among Global South countries to shape international AI standards rather than simply adopting rules developed elsewhere.
Get Newsletter Updates
Enjoying our column?
Subscribe to our specialised **Tech Pulse** feed to receive fresh reports and analyses directly in your inbox.
