Child-Centered AI Policy
Public commitments, each with an accountable owner.
Our definition of a child, across childhood and adolescence.
Effective 01 Jan 2026, reviewed at least annually.
Before We Begin
Children do not choose the AI that shapes their learning. We choose it for them with every product decision we and our partners make. This policy holds us to that responsibility in the open.
It governs everything we build today and design tomorrow, proactively outpacing regulation and strictly protecting a child's trust from the start.
An EdTech company stands on its written commitments, not its claims. This document is that commitment.
Purpose
This policy defines how we design, build, and operate AI-enabled learning products to ensure every child is safer, better served, and more capable. It translates international children's rights and child-centered AI principles into concrete commitments across our products, partnerships, and daily decisions.
Publishing these promises in the open allows parents, educators, regulators, and children to hold us directly accountable.
Scope & Definitions (applies to)
- Every product and feature we build or operate that is used by, involves data about, or is developed for children under 18.
- All deployed AI systems, whether built in-house, licensed, or integrated through partners.
- All internal departments—including product, engineering, data, content, design, and support—along with external partners and vendors.
- All operational regions, enforcing the highest applicable standard even where local laws differ.
Definitions
Our Approach
This policy is not a marketing document. It is grounded in internationally recognized principles for children's rights, safety, and development, and benchmarked against the most rigorous publicly available criteria for evaluating whether an EdTech or AI product is genuinely safe, evidence-based, designed for children, scalable, and inclusive — not simply marketed to their parents.
We used those principles and criteria to test our own assumptions, not to decorate this document. Where our existing practice fell short of what they ask for, this policy commits us to closing that gap, not to writing around it.
Our Ten Commitments to Child-Centered AI
Everything below covers governance, safety, data, fairness, transparency, rights, wellbeing, inclusion, AI literacy, and our enabling environment.
- 1. Ensure Regulatory Frameworks, Oversight and Compliance.
- 2. Ensure Safety for Children
- 3. Protect Children's Data and Privacy
- 4. Ensure Non-Discrimination and Fairness
- 5. Provide Transparency, Explainability and Accountability
- 6. Respect Human and Child Rights Through Responsible AI Practice
- 7. Support Children's Best Interests, Development and Well-Being
- 8. Ensure Inclusion of and for Children
- 9. Prepare and Skill Children for Present and Future Developments in AI
- 10. Create an Enabling Environment for Child-Centered AI
We govern our AI like the responsibility it is accountable, documented, and open to scrutiny.
Compliance is the floor, not the ceiling. We build governance for child-centered AI into how we operate, anchored in international guidance and the law of every market we serve.
We commit to:- Comply with applicable data protection, child-safety, and AI-specific regulation in every jurisdiction (e.g. GDPR, COPPA, EU Digital Services Act), applying the strictest standard where local laws differ.
- Maintain a named, accountable owner for child safety and responsible AI, with authority to delay or withdraw any feature that does not meet this policy.
- Document the pedagogical and ethical rationale behind our solutions — a living explanation of how each product benefits children and how that benefit is measured.
- Subject our practices to regular internal review and, as we mature, to external assessment frameworks and independent certifications.
- Be transparently and legally incorporated, and make our governance structure available to regulators, partners, and the public on request.
Safety is designed in before a child ever sees a feature, not patched in after.
Anything a child can see, hear, or talk to is held to safety-by-design from the first prototype, with human judgement kept firmly in the loop.
We commit to:- Apply safety controls and guardrails to every generative or interactive AI feature, with human review of AI-generated content before and after release.
- Moderate user interactions and user-contributed content before children are exposed to it, combining automated detection with human oversight.
- Design explicitly against manipulation and addictive engagement patterns, with extra safeguards on any conversational AI
- Maintain a cybersecurity program covering threat detection, vulnerability management, and incident response.
- Provide a clear, accessible channel for children, parents, and educators to report a safety concern, with a defined response time and escalation path.
Children's data is held in trust — never sold, never used to target them.
We apply privacy by design and by default, treating a child's data as something we are entrusted with, not a commercial asset to exploit.
We commit to:- Collect only the data needed to deliver the learning experience, requesting consent in clear, age-appropriate language including parental consent wherever required.
- Never sell children's personal or non-personal data, and never use children's data for advertising profiles or to target children with ads.
- Run no third-party advertising within any child-facing experience.
- Publish a privacy policy that plainly explains what data is collected, why, how long it is kept, and who it is shared with.
- Share data with third parties only under a documented data-sharing agreement and a clear lawful basis, never by default.
- Be transparent about the origin and provenance of datasets used to train or fine-tune our AI systems.
- Comply with applicable data-protection regimes including GDPR and COPPA, and review our practices whenever those frameworks evolve.
AI should not repeat the world's biases onto the next generation.
Capable AI systems can quietly encode the biases of their data and their designers. We test for that, rather than assume it away.
We commit to:- Test our products and the models behind them for differential outcomes across gender, disability, ethnicity, language, and socio-economic background, and correct any disparities.
- Design with the children mainstream EdTech most often leaves out.
- Build gender equity into product design and content, avoiding stereotyped representation.
- Involve a diverse range of voices including educators and communities from the contexts we serve in design and review.
- Designed especially for children with disabilities, refugees & migrants, out-of-school children, ethnic minorities, girls in unequal-access contexts.
Children, parents, and teachers deserve a real answer, not a black box.
Understanding what our AI is doing, and why, should not require a law degree. We owe a plain-language explanation to the people actually affected.
We commit to:- Clearly disclose when content, feedback, or a recommendation a child receives was generated or shaped by AI.
- Offer plain-language explanations of how our products personalize content, assess progress, or make recommendations — written separately for children, parents, and teachers.
- Keep a human accountable for every consequential AI decision, with a named internal owner for each AI feature.
- Provide an accessible way for a parent, educator, or child to question or appeal an AI-driven decision, or flag output that looks wrong.
- Share evaluation results and safety performance with partners, frameworks, and communities — not only with regulators.
Every product decision is measured against a child's rights, not just our roadmap.
We treat a child's fundamental rights — to protection, privacy, participation, and non-discrimination — as a design specification, not a poster on the wall.
We commit to:- Treat the rights to education, protection, privacy, non-discrimination, participation, and freedom of expression as design requirements for every product.
- Extend human-rights due diligence across our full AI value chain, including third-party models, datasets, and vendors.
- Avoid AI uses that go beyond legitimate pedagogical need into surveillance or profiling of children.
- Apply the same rights-respecting standard to AI systems used by children in administrative, assessment, or monitoring contexts.
We measure success by what a child learns and how they feel, not by time on screen.
Our products exist to help a child develop, not to maximize the minutes they spend with us. Those are not the same goal, and we design for the first one.
We commit to:- Ground every product in documented learning objectives and pedagogical design, reviewed by subject-matter experts and aligned with recognized educational standards.
- Generate and act on real evidence of learning impact through studies and evaluations — not treating engagement metrics as a stand-in for learning.
- Design for healthy use: avoid compulsive engagement loops, encourage natural breaks, and support rather than substitute for the role of caregivers and educators.
- Support the whole child: critical thinking, creativity, collaboration, and social-emotional development alongside subject-matter content.
- Continuously collect feedback from students, teachers, and parents with a documented process for turning that feedback into product changes.
If a child cannot reach it, afford it, or use it, it is not finished.
A child-centered AI policy means little if whole groups of children cannot reach or use the product in the first place.
We commit to:- Design and test for accessibility against the WCAG 2.1 AA standard, including testing with assistive technologies and people with disabilities.
- Build for the infrastructure children actually have: low-bandwidth and offline functionality, compatibility with low-cost devices, and reasonable pricing.
- Localize meaningfully — language, curriculum alignment, and cultural representation — rather than simply translating text.
- Create real opportunities for children, families, and educators to shape our products, treating them as participants in design and feedback.
We want every child to leave smarter about AI than they arrived.
Children should come away from our products better equipped to understand, question, and use AI — not more dependent on it without understanding it.
We commit to:- Help children recognize when they are interacting with AI, and build age-appropriate understanding of what AI can and cannot do.
- Build digital literacy, critical thinking, and healthy skepticism toward AI-generated content directly into the learning experience.
- Equip parents and educators with guidance and training materials to support children's safe and effective use of AI-enabled learning tools.
- Track and report which future-facing skills our products help build alongside core academic outcomes.
Commitments only matter if they are lived, reviewed, and renewed.
Child-centered AI has to be an organization-wide practice, not a policy filed away after launch.
We commit to:- Embed a child-rights and safety review into our product development lifecycle so new AI features are assessed before launch, not after a problem emerges.
- Provide regular training on child rights, child safety, and responsible AI practice for every team that touches a child-facing product.
- Actively engage with the wider ecosystem working on this issue so our practice keeps pace with an evolving field.
- Treat this policy as a living document, reviewed at least annually and sooner whenever regulation, technology, or our own products create new risk or opportunity for children.
- Maintain visible, accountable ownership of this policy at the leadership level, and resolve any conflict between a business goal and a child's best interests in the child's favor.
Governance, Implementation and Accountability
A policy is only as strong as the structure that enforces it. We commit to:
Ownership
Review Process
Training
Vendors
Reporting Channels
Tie-breaker
- Ownership. The accountable executive owner for child safety and responsible AI has the authority to delay or stop features that might not meet the policy.
- Review process. Embedding a child-rights and safety review into our product development process so new AI features are assessed before launch rather than after an incident.
- Training. Training teams responsible for child-facing products—including content, design, marketing, and support—on the commitments in this policy.
- Vendors. Ensuring third-party vendors that we work with meet our standards, documented in vendor agreements.
- Reporting channel. Providing a clear, accessible channel for children, parents, and educators to raise a concern, with a committed response time and an escalation path that reaches the accountable owner.
- Tie-breaker. Revisiting possible conflicts or standard overlaps between business objectives and child-centered interests, ensuring every resolution is logged for compliance.
A Living Policy
- Review cadence. Reviewing this policy frequently to address new regulations, technologies, or opportunities that support children's safe development.
- Version history. Publishing material updates with a visible version history so anyone relying on this policy can see what has changed and why.
- Benchmarking. Continuing to compare our practice against the best available external frameworks and emerging standards for child-centered AI and EdTech.
Closing Commitment
Children do not get a vote in how AI is built, but they live with the consequences of how we build it. This policy is our answer to that responsibility: not a marketing claim, but a working document our products, our partnerships, and our own teams are measured against.
We will not always get every detail right on the first try. We commit to saying so plainly when we do not, and to fixing it quickly when we are shown a better way.
See something that does not match this policy?
Whether you are a parent, an educator, or a child using one of our products — every report reaches a named owner and receives a response.
