Tech giants race to embed AI in schools globally, raising concerns over critical thinking

In the span of a few weeks last fall, three major technology companies announced plans to flood schools with artificial intelligence. Microsoft would supply AI tools and training to over 200,000 students and educators in the United Arab Emirates. Days later, a financial services company in Kazakhstan struck a deal with OpenAI to distribute ChatGPT Edu to 165,000 educators. Then Elon Musk's xAI announced something even larger: a tutoring system powered by its Grok chatbot for more than a million students across thousands of schools in El Salvador. The pattern was unmistakable. Governments worldwide, often encouraged by American tech firms, were racing to embed generative AI into classrooms and universities, betting that the technology would personalize learning, save teachers time, and prepare young people for an AI-driven economy.

The pitch from Silicon Valley was familiar. Chatbots could generate human-like emails, create quizzes, analyze data, and write code. They could democratize education and level the playing field for students in underserved regions. But this narrative had echoed before. Years ago, the same companies had promised that laptops would revolutionize learning. The One Laptop per Child initiative, a global effort to put computers in schools across the developing world, was studied extensively in Peru and elsewhere. The results were disappointing. The programs consumed resources without improving students' cognitive abilities or academic outcomes. Now, as tech enthusiasts made nearly identical arguments about AI and educational equity, child advocacy organizations like UNICEF were sounding alarms. Steven Vosloo, a digital policy specialist at UNICEF, warned that "unguided use of AI systems can actively disqualify students and teachers." The stakes, he suggested, were higher than wasted money. They involved the fundamental development of young minds.

Research was beginning to surface troubling patterns. A study by Microsoft and Carnegie Mellon University found that popular AI chatbots could diminish critical thinking. The bots produced errors and misinformation with an air of authority, making them especially dangerous in educational settings where students were still learning to evaluate sources. Teachers were already reporting widespread cheating assisted by AI. In the United States, where individual states and school districts controlled curriculum decisions, major systems were moving fast. Miami-Dade County Public Schools, the third-largest district in the nation, deployed Google's Gemini chatbot to over 100,000 high school students. Broward County, the sixth-largest district, introduced Microsoft's Copilot to thousands of teachers and staff. Internationally, Microsoft partnered with Thailand's Ministry of Education to offer free AI skills courses to hundreds of thousands of students, then committed to training 150,000 teachers. OpenAI promised to make ChatGPT available to public school teachers across India.

But two small nations were attempting something different. Estonia, a Baltic country with a history of technological innovation, launched a national initiative called "AI Leap" rooted in a troubling discovery: over 90 percent of the country's high school students were already using chatbots like ChatGPT for homework. Some were delegating entire assignments to the machines. Rather than simply distributing tools, Estonia pressed American tech giants to adapt their products to local educational needs. Researchers at the University of Tartu worked with OpenAI to modify ChatGPT for Estonian schools so that it would respond to student queries with questions rather than direct answers—a design choice meant to preserve critical thinking. The AI Leap program, introduced this school year, aimed to teach educators and students about the uses, limits, biases, and risks of AI tools. Ivo Visak, executive director of the AI Leap Foundation, described it as "critical literacy in AI." The goal was not to ban the technology or distribute it uncritically, but to help people understand that these tools "can be useful—but at the same time these tools can cause much damage."

Iceland pursued a parallel path. This school year, it launched its own national AI pilot, with several hundred teachers experimenting with Google's Gemini or Anthropic's Claude for tasks like lesson planning. Crucially, students were not using the chatbots yet. Thordis Sigurdardottir, Iceland's director of Education and School Services, explained the caution: "If you are using less of your brain power or critical thinking—or what makes us more human—that definitely is not what we want." Two teachers participating in the pilot, Tinna Arnardottir and Frida Gylfadottir, found the tools genuinely useful for creating engaging lessons faster. Arnardottir used Claude to build a career exploration game for her business students; Gylfadottir loaded vocabulary lists and had the chatbot generate word games and matching exercises. But both teachers were vigilant. Gylfadottir checked the accuracy of every AI-generated game before students saw it. Both worried that their students were already becoming dependent on AI outside school, blindly trusting outputs without verification. This concern made them more determined to teach students how to think critically about and with these tools, not simply to use them.

The broader landscape remained uncertain. Teachers had few rigorous studies to guide them in deploying generative AI. Researchers were only beginning to track the long-term effects of chatbots on adolescents and school-age children. Drew Bent, an education leader at Anthropic, acknowledged the moment: "Many institutions are experimenting with AI. We are at a point now where we need to make sure these things are backed by results and figure out what is working and what is not working." The question facing educators worldwide was whether the technology companies pushing these tools would adapt them to serve schools' actual needs, or whether schools would simply become markets for products designed elsewhere. Estonia and Iceland were testing whether a third path existed—one where critical thinking remained central, where technology served human judgment rather than replacing it, and where the lessons of previous tech failures were actually learned.