A new study from the Massachusetts Institute of Technology has intensified debate over the use of generative artificial-intelligence tools in classrooms. Researchers found that college students who drafted essays with ChatGPT exhibited 47% less brain activity, as measured by functional-MRI scans, than peers who wrote unaided. Neuroscientists Daniel Amen and Terry Sejnowski said the findings add to evidence that prolonged reliance on large-language models could dull critical-thinking skills if left unchecked. The research lands as surveys show the technology is already ubiquitous. A 2024 Pew poll reported that 26% of U.S. teenagers had used ChatGPT for schoolwork, double the previous year, while an internal Harvard survey indicated almost two-thirds of its undergraduates tapped AI tools weekly by last spring. Faculty responses range from reverting to handwritten exams to incorporating AI literacy modules, but many professors acknowledge they are struggling to keep pace. Student-data security is emerging as a parallel concern. Education-technology platforms increasingly route homework, essays and personal details through commercial AI systems, yet the main U.S. privacy statute, the 1970s-era Family Educational Rights and Privacy Act, has never been enforced, according to the Center for Democracy and Technology. Last year’s PowerSchool breach, which exposed thousands of records, underscored the sector’s vulnerabilities. Governments are moving in divergent directions. Australia this week unveiled a pilot program that will embed artificial-intelligence coursework across selected K-12 schools, aiming to prepare roughly one million students for future jobs. By contrast, some U.S. districts have tightened rules on chatbot use until clearer safeguards are in place. Policymakers on both fronts cite the MIT findings as fresh impetus to balance AI’s instructional benefits with its cognitive and privacy risks.