India’s Digital Public Infrastructure for Children’s AI Wellbeing
Documentation • Architecture • Packages • Get Started • Contributing
CHAISE (Children, AI and Safe Environments for Growth) is a comprehensive Digital Public Infrastructure (DPI) framework designed to ensure children’s safety and wellbeing in the age of AI and social media.
Children deserve:
- ✅ Age-appropriate digital experiences
- ✅ Protection without over-restriction
- ✅ Safe spaces to learn and explore
- ✅ Transparency and accountability from platforms
Modern child safety requires smarter protection through identity verification, intelligent policies, real-time protection, and comprehensive transparency across digital platforms.
Children today face unprecedented challenges in digital spaces:
| Era | Risk Type | Protection Needed |
|---|---|---|
| 🌐 Static Web | Exposure to harmful content | Content filters, parental controls |
| 💬 Social Web | Cyberbullying, grooming, stranger danger | Safe interaction rules, adult oversight |
| 🤖 Algorithmic Feeds | Addiction loops, behavioral manipulation | Healthy boundaries, transparent algorithms |
| ✨ Generative AI | Synthetic content, personalized threats | Identity verification, real-time monitoring |
The Challenge: Current solutions are fragmented, reactive, and don’t address the architectural nature of these risks.
CHAISE provides an end-to-end safety pipeline that operates at every layer of the digital experience:
graph LR
A[Identity Graph] --> B[Age Token System]
B --> C[Policy Objects]
C --> D[On-Device Cache]
D --> E[Content Curation]
E --> F[Audit Logging]
- 🔐 Identity Graph - Verifiable relationships with cryptographic authority proofs
- 🎫 Age Token System - Privacy-preserving age verification without exposing PII
- 📋 Policy Objects - Natural language policies converted to enforceable rules
- 💾 On-Device Cache - Offline functionality with encrypted storage
- 🔍 Content Curation - Real-time filtering through service provider integration
- 📊 Audit Logging - Comprehensive tracking with selective disclosure
CHAISE implements a defense-in-depth architecture with 8 specialized packages:
- Package 1 - CAATS (Child Attribute & Age Token Service)
- Issues privacy-preserving age tokens
- Supports multiple digital ID providers
- Package 2 - CRGS (Child Relationship Graph Service)
- Manages guardian, teacher, and agent relationships
- Provides authorization proofs
- Package 3 - NLPPS (Natural-Language Policy Authoring & Store)
- Natural language to machine-enforceable rules
- Child voice and appeal mechanisms
- Package 4 - CCCL (Confidential Content Certification & Labeling)
- Privacy-preserving content certification
- Third-party certifier ecosystem
- Package 5 - SPEC (Server-side Policy Evaluation & Curation Pipeline)
- Stackable curation engines
- Provider-neutral enforcement
- Package 6 - OCVL (On-device Contextual Curation & Dynamic Control)
- Context-aware second-factor decisions
- Coaching and soft interventions
- Package 7 - RISIR (Real-time Interaction Safety & Incident Response)
- Grooming detection and prevention
- Live interaction monitoring
- Package 8 - EPT-EC (Evidence, Provenance & Traceability)
- End-to-end audit trails
- Selective disclosure for authorities
- Age bands instead of exact dates of birth
- On-device processing for sensitive data
- Scoped pseudonyms, no global tracking
- Separation of safety from advertising analytics
- Clear explanations and coaching
- Appeal and negotiation mechanisms
- Graduated autonomy by age
- Balance between safety and learning
- Server-side + device-side protection
- Multiple certification engines
- Fallback modes and resilience
- No single point of failure
- Pluggable engine ecosystem
- Multiple ID providers supported
- Federation-ready architecture
- Jurisdiction-aware policies
- Complete Specification - Full technical specification (v1.0)
- Architecture Guide - Detailed architecture documentation
- Diagrams - Relevant diagrams
- Implementation Guide - How to implement CHAISE
- API Reference - Package APIs and interfaces
- Use Cases - Real-world implementation scenarios
- Trouble Shhoting - Troubleshoot
- Contributing - FAQ
- FAQ - FAQ
- Review the specification - Understand the 8-package architecture
- Identify integration points - Determine which packages apply to your platform
- Implement core packages - Start with CAATS, CRGS, and NLPPS
- Add certification - Integrate with CCCL for content labeling
- Deploy curation pipeline - Implement SPEC and OCVL
- Enable real-time safety - Add RISIR for live interactions
- Implement traceability - Deploy EPT-EC for audit and compliance
- Understand the framework - Review architecture and principles
- Define jurisdiction requirements - Customize policies for local context
- Establish governance - Set up certifier registries and standards
- Enable federation - Support multiple providers and engines
- Monitor compliance - Use EPT-EC for oversight and evidence
- Study the specification - Analyze the technical approach
- Evaluate effectiveness - Research impact on child safety
- Improve algorithms - Develop better detection and curation engines
- Share findings - Contribute to the knowledge base
- Collaborate - Work with implementers and policymakers
We welcome contributions from:
- Developers - Implementation, SDKs, tools
- Researchers - Safety algorithms, evaluation frameworks
- Policymakers - Regulatory guidance, compliance frameworks
- Advocates - Child safety expertise, user research
- Read our Contributing Guidelines
- Check Open Issues
- Fork the repository
- Create a feature branch
- Submit a pull request
- Age-appropriate content curation
- Teacher-defined policies during school hours
- Safe exploration with learning goals
- Real-time grooming detection
- Context-aware content filtering
- Parent-teen negotiation workflows
- Interaction safety in multiplayer
- Time and context limits
- In-game purchase controls
- AI-generated content certification
- Personalized safety without tracking
- Transparent recommendation systems
CHAISE is built on strong security and privacy foundations:
- Zero-knowledge proofs for age verification
- Confidential computing for content certification
- End-to-end encryption for sensitive data
- Minimal data retention by default
- Selective disclosure for authorities
- Privacy-preserving audits with hash-chained logs
- Support for multiple national digital identity systems
- Privacy-preserving credential presentation
- Age band verification without exposing DOB
- Independent third-party certification
- Stackable certification standards
- Attestable execution environments
- Pluggable algorithm marketplace
- Auditable and versioned
- Provider-neutral implementations
- Educational platforms
- Social media services
- Gaming platforms
- Content streaming services
- ✅ v1.0 Specification released
- 🔄 Reference implementation of core packages
- 🔄 SDK development for major platforms
- 🔜 Certifier onboarding and registry
- 🔜 Curation engine marketplace
- 🔜 Multi-platform pilots
- Blog: ISPIRT Foundation
- Community: Join Discussion
This specification and reference implementation are released under [MIT].
For commercial implementations and certifications, please contact ISPIRT Foundation.
CHAISE is developed by the ISPIRT Foundation DEPA Team with contributions from:
- Child safety experts and advocates
- Platform providers and developers
- Academic researchers
- Policymakers and regulators
- Parents and educators
Special thanks to all contributors who have helped shape this critical infrastructure for children’s digital wellbeing.
ISPIRT Foundation | DEPA Team
- Email: shyam or sunu or subodh dot sharma @ispirt.in
- Website: https://ispirt.in
- GitHub: https://github.com/iSPIRT
Protecting Children in the Age of AI
Because every child deserves a safe digital future