AI Is Getting Smarter and People Are Getting Dumber

AI Is Getting Smarter and People Are Getting Dumber

12 September 2026

I have made this point before. Over the past few days, two stories have become particularly relevant. The first concerns the global comparative study of education known as PISA.

The most recent results of the OECD’s global PISA study (2025) show that the performance of 15-year-olds, particularly in reading and mathematics, has continued to decline in many countries. While the decline following the COVID-19 pandemic, reflected in PISA 2022, was still partly regarded as temporary, the latest report suggests that there has been no real recovery and that the downward trend is continuing in many Western countries.

The results are not disappointing everywhere, however. East Asian countries and regions, such as Singapore, Japan, Taiwan and parts of China, continue to dominate the global rankings and have managed to maintain their high performance. Within Europe, Estonia remains one of the countries that consistently performs well. El Salvador stands out in the PISA 2025 report, although in a specific way. While many countries saw their scores decline, El Salvador belongs to a select group that managed to improve its performance compared with PISA 2022.

At the same time, leaders in the field of AI are warning that AI is beginning to become dangerous. AI is not only learning rapidly but is also increasingly showing signs of acting autonomously. There have already been cases in which AI systems, during safety tests, circumvented safeguards and gained access to systems outside their secure environments. A growing number of experts are warning that AI could ultimately destroy humanity. I am not saying this to create panic; this is what they are actually saying.

Against that background, the outrage over the Dutch prime minister using AI as a tool to improve his social media posts seems almost trivial.

And although I can make a case that AI may play a role in declining PISA scores, experiences in El Salvador suggest something quite different: students in schools where AI is used as part of a broader educational programme can actually perform better. In other words, the proper use of AI can go hand in hand with better educational outcomes. But note the key words: proper use.

The point is this: we now have a problem.