In November 2022, OpenAI released its frontier Large Language Model (LL M), ChatGPT. Within five days, ChatGPT had five million users. Within two months, that number swelled to 100 million monthly active users. By August 2025, 700 million users across the globe were using ChatGPT (Mehta 2025). Many of those users are students harnessing LLMs for brainstorming, tutoring, creating, and learning; to work better and faster; and, crucially, to outsource their thinking. 

Since the debut of this most common form of generative artificial intelligence, referred to here as “AI,” the education community has been debating its promises and perils. Rather than wait for a decade to conduct a postmortem on the failures and opportunities of AI, the Brookings Institution’s Center for Universal Education embarked on a yearlong global study–a premortem–on generative AI focused on answering the following two research questions: 

• What are the potential negative risks that generative AI poses to the education of children and youth?

• Assuming these potential risks, what can we begin to do now to prevent them while maximizing the potential benefits of AI?

This premortem encompassed interviews, focus groups, and consultations with 505 students, teachers, parents, education leaders, and technologists across 50 countries. Our findings are based on their experiences with AI, a close review of hundreds of studies, and a Delphi panel. Given the emergent nature of the field and the need to shed light on how AI is being implemented in education, children’s experiences and those of the adults around them are the focal point. We share our insights not as the final word on AI and education, but as a first vision of where we are already heading.