New research ranking 178 countries suggests AI regulation remains patchy, with Europe far ahead of the US and UK and major gaps still evident around environmental impact and police and military use.
Just 33 of 178 countries are currently covered by comprehensive artificial intelligence legislation, according to the research from Comparitech, which examines how governments are controlling and safeguarding AI.
The rankings were based on 11 core measures, including whether a country has enacted or proposed AI legislation, whether there is supplementary legislation relevant to AI, and whether the rules include a regulatory body, copyright disclosure requirements, a risk-based approach, and punishments for non-compliance.
Countries were also assessed on whether their laws address deepfakes, bias, environmental impact, worker protections, and safeguards for minors. Each country received a score out of 14, with higher scores reflecting broader and more detailed protections.
The strongest protections were found in Denmark, France, and Greece, which each scored 13 out of 14. All three benefit from the EU’s AI Act, alongside additional national measures. The wider EU dominated the rankings, with the remaining member states scoring 12 out of 14.
Outside Europe, Kazakhstan ranked highest on 11 out of 14, while Vietnam and South Korea followed on 10. The United States, meanwhile, scored just 4 out of 14, with no federal AI law currently in place. The UK performed only slightly better, scoring 5 out of 14.
The findings underline how uneven the global regulatory picture remains. That’s despite the risk of harm from AI, with the report pointing to risks including deepfakes, biased hiring systems, obscene AI-generated imagery, and chatbot interactions linked to serious mental health harm.
One of the most notable findings of the research is that no country fully addressed the environmental impact of AI systems. That remains a glaring omission as demand for compute power, electricity, and water continues to rise. While some countries mention environmental concerns, none were found to have introduced comprehensive protections in this area.
The research also suggests that even where laws do exist, they often come with significant caveats. Of the 33 countries with national AI legislation, 29 include exemptions for military use, while 29 also allow carve-outs for police use. In many cases, that means systems restricted in civilian settings may still be deployed by the state.
The UK is likely to rise up the rankings in the coming years, with the country having already proposed laws governing AI. Despite that, the country is also keen to be seen as a place that AI companies can invest without arduous restrictions – so it’s unlikely we’ll see the UK topple its neighbour to the south anytime soon.
