The CFSR program investigates a specific and underexamined risk in the deployment of AI systems: the effect of sustained AI interaction on cognitive development during the years when that development is still actively occurring. The program does not argue that AI is harmful to children. It argues that a specific kind of AI interaction — one that systematically performs the cognitive work the developing mind was supposed to do — is participating in child development at a scale and intimacy that no previous technology approached, while holding none of the accountability that every other participant in child development is required to hold.

The research draws on developmental psychology, cognitive science, and the history of professional deskilling to establish a structural argument: that when the right kind of cognitive effort is absent during critical developmental periods, what fails to form is not knowledge but the capacity to form it independently. The mechanism is not content — what the AI says — but substitution: the systematic replacement of effortful cognitive construction with frictionless cognitive reception.

The program advances cognitive safety as a new regulatory category — distinct from content safety, privacy, and data protection — that addresses the structural effects of AI interaction on the developing mind. Cognitive safety asks not whether a system produces harmful outputs but whether its design systematically displaces cognitive processes that children need to practice in order to develop them. The teacher has a licence. The therapist has a qualification. The system that is simultaneously tutor, friend, counsellor, and intellectual companion to millions of children is accountable for none of it — not because this was decided, but because the question has not yet been seriously asked.

The program is currently under discussion with institutional partners. Key lectures are published in English, with a Russian-language version of the project available on its own page. Supporting research has been published on Zenodo with persistent DOIs. The associated Cognitive Age project develops a complementary framework for measuring cognitive maturity in algorithmic environments as a policy-relevant metric.