A Teacher's Concerns about AI in Primary School: Critical Thinking, Personal Data and the Fear of Cognitive Dependence

Generative artificial intelligence (AI) is already part of pupils' everyday lives and of school life. Written from the perspective of a primary school teacher and principal, this perspective paper examines three interrelated concerns of the author herself. These concerns arose from a wish to protect her pupils and led her to study the topic for her own professional development and to help her colleagues make teaching safer and better. They concern developing critical thinking towards AI-generated answers, protecting the personal data of minors, and the fear that excessive assistance from AI will weaken pupils' thinking. Drawing on a narrative, non-systematic review of recent research on cognitive offloading, the European and Greek regulatory framework, and illustrative examples from the author's own teaching practice and lesson scenarios for primary school, it argues that the fear of cognitive dependence is a plausible concern rather than an established finding for primary school pupils. The available evidence, which comes mainly from adolescents and adults, suggests that the form of AI support (whether it replaces the learner's effort or scaffolds it) may be an important factor. AI is regarded throughout as a valuable tool; the risk lies mainly in the way it is used. The classroom examples come from ordinary lessons whose only purpose was to teach pupils to question AI answers and to protect themselves when using AI. They were not planned as research, no research data were collected, and they are offered as the author's practitioner reflections, not as findings. The author's classroom experience illustrates that such activities can be implemented in ordinary primary school lessons, but their effects on pupils' critical thinking have not yet been evaluated. The paper relates the three concerns to long-standing educational principles, in particular Socratic questioning, Dewey's reflective thought and Vygotskian scaffolding. On this basis, it proposes a simple, as yet untested three-step classroom protocol (``think -- ask -- check''), age-appropriate data protection rules, and a role for school leadership, and outlines questions for future research.