Child Safety & AI Governance Framework

1. Child Safeguarding Policy

The principles that put children's safety and privacy first.

• Welfare Primacy: The psychological safety, physical well-being, and digital privacy of minors represent the supreme design parameters of all IWA systems.
• Differentiated Age Control: User interfaces and platform controls must dynamically scale based on the age cohort (10–12, 13–15 and 16–17) to respect developing privacy expectations and distinct parental validation rules.
• Adult-Child Boundaries: The platform architecture blocks unmonitored adult-to-child or child-to-child interactive communication loops, keeping user text reflection boxes strictly private.
• Explicit Reporting Gateways: Highly visible, clear, and age-appropriate reporting buttons must be available for children to trigger immediate manual review if they encounter distressing topics.
• Target Safety Boundaries: This policy regulates systems monitoring for: platform bullying, abuse disclosures, self-harm indicators, external physical threats, exploitation risk patterns, or inappropriate adult contact attempts.
• Algorithmic Separation: AI safety alerts are categorized as informational notifications requiring human confirmation; they are never displayed as formal clinical or psychological evaluations of a user.

2. AI Safety & Responsible AI Policy

The limits we place on AI, so it supports children and never judges them.

• Clear Product Purpose Only: AI components must strictly serve the direct educational and character-building features of the platform.
• Minimization & Pseudonymization: AI systems must operate strictly on stripped, decoupled text inputs. No identifying details or metadata may be exposed to the AI gateway.
• Human Oversight Integration: AI safety flags serve as operational triggers for immediate human review; algorithmic responses must never make automated, definitive decisions regarding a child’s access.
• Absolute Profiling Ban: AI components are strictly prohibited from diagnosing, psychologically assessing, labelling, ranking, or predicting children.
• Removal of Behavioral Inference: All legacy processing for tracking “willingness” or “engagement text profiling” is formally terminated. IWA restricts processing to safety moderation only, eliminating behavioral tracking.
• Fact Separation: Algorithmic evaluations must never be rendered or stored as objective facts regarding a child's inherent character, personality traits, or mental health status.
• Commercial Restrictions: Under no circumstances will any data input generated by a child be utilized for behavioral advertising, promotional targeting, or commercial profiling.

3. How this works in practice

• Accounts are created only by a parent or a school; students and teachers cannot sign themselves up.
• We collect a first name, nickname or student number, an age group and a system username — never surnames, photos, videos or voice recordings.
• There is no chat between students, no public profile and no contact from strangers. Reflections are private by default.
• Before any reflection is safety-checked, names, emails and phone numbers are masked. The AI receives only the text and a one-time random reference.
• Safety alerts carry no names. A name appears only when an authorised adult — the parent, the school's Super Admin or an IWA Admin — opens the alert while signed in.
• Every student has a “Report a concern” button and a “Need help now?” list of helplines on every page. Concerns a child raises themselves are not sent to a parent automatically, in case the worry involves home.