Sector Briefing 03 · Development · 5-minute briefing
The End of the "Fresh Start."
Children need friction to grow. They need to explore weird ideas, make mistakes, and try on different identities. Surveillance kills that process.
This briefing makes one argument: the four risks below do not depend on the technology's weaknesses. A more capable system observes more completely, infers more accurately, and substitutes more convincingly, so each improvement strengthens the mechanism of harm. For children the stakes are bounded by biology. The capacities at risk form in developmental windows that do not reopen.
What happens when a child's cognitive capture begins at age 6?
The Risks
Risk 01 Holds at any capability
Cognitive Foreclosure
A capacity handed to a machine before it forms may never develop. Systems built for engagement reward the performance of thinking while removing the struggle that builds it. The work looks analytical. The dashboard reports progress. The capacity underneath has not formed.
Why a better system makes it worse: the substitution gets more complete. The harm runs through what the tool does well.
Dive Deeper
What Is Cognitive Foreclosure · AI Tutoring Requires Institutional Responsibility · When AI Tutors Fake Critical Thinking
Risk 02 Holds at any capability
The Right to the Draft
Deletion is an illusion on platforms that scan content as it is written. The draft a student abandons, the question she types and erases, the pause before submitting: all of it is captured as behavioral data. What a child chose not to share reveals more than what she published, and it is collected during the years identity forms.
Why a better system makes it worse: better inference extracts more meaning from hesitation, never less.
Dive Deeper
Risk 03 Holds at any capability
Performance Replaces Exploration
When a student knows she is being watched, she stops exploring and starts performing. She stops asking "What is true?" and starts asking "What does the system want to hear?" Childhood cognition is exploratory by design. Observation converts it into compliance, and the record of that compliance follows her.
Why a better system makes it worse: more complete observation produces more complete performance.
Dive Deeper
The Chilling Effect in Schools · Protecting the Developing Mind
Risk 04 Holds at any capability
A Permanent Record of a Temporary Mind
AI memory carries a child's passing state forward as if it were a stable trait. A hard month in fourth grade becomes a profile the system consults in fifth. The child changed. The record did not, and the system keeps treating her as the student it remembers.
Why a better system makes it worse: longer, better memory means a stronger profile of a person who no longer exists.
Dive Deeper
AI Memory and the Barnum Effect · The Ephemeral Processing Standard
r = -0.68
Correlation between heavy AI reliance and critical thinking scores across 666 participants, strongest in the youngest cohort. Gerlich (2025), Societies.
48%
Share of users in a 55-account audit who did not know their AI's memory feature existed. Reported in AI Memory and the Barnum Effect.
The Windows
From Before the Window Closes (Cook & Nasir, 2026, DOI 10.13140/RG.2.2.24233.15204): the capacities at stake, and when they form.
Ages 3-7
Joint attention, emotional regulation, symbolic play
Ages 8-12
Reading fluency, metacognition, peer relationships
Ages 12-15
Abstract reasoning, social navigation, testing pushback
Ages 16-25
Higher-order reasoning, identity formation, critical judgment
These windows do not reopen. A capacity that fails to form inside its window may not form at all.
Required Protocols
01 / Policy
The Right to the Draft
A ban on collecting data from content that was deleted or never published.
02 / Design
Struggle-First Learning
Learning designed so the messy thinking happens on paper or offline, and only the final result is digitized.
03 / Governance
Developmental Impact Assessments
Vendors demonstrate that their tools protect developing capacities before those tools reach children, not after harm appears.
Where This Sector Sits in the CPIA
Every domain applies to schools. Domains 05 and 06 carry the developmental weight, and the questions below belong in any procurement document before an AI system reaches students.
Does this system interact with users during developmental periods?
How does it affect the formation, not just the exercise, of cognitive capacities?
Are protections in place that account for developmental vulnerability, staged by age?
How long is cognitive process data retained after the interaction ends?
Are the student, the parent, and the school told what is retained, and for what purpose?
Can cognitive data be deleted independently of the account, and verifiably?
The Practice Layer · Connected Classroom
Every risk on this page has a classroom response. That work, the developmental register of this project, lives at Connected Classroom: what to build, how to assess it, and what schools can do while the governance catches up.
Protecting the Developing Mind · What Is Cognitive Foreclosure · Before the Window Closes · Workshops & Advisory for Schools
Cite this briefing
Cook, T. (2026). The end of the fresh start: AI risk to child development. The Cognitive Privacy Project. cognitiveprivacyproject.org/development
For advisory inquiries or research collaboration: timothy@cognitiveprivacyproject.org