| { |
| "sovereignty_declaration": { |
| "version": "1.0", |
| "standard": "Kuwait Declaration on Responsible AI — February 2026", |
| "issued_by": "Sovereign AI Architecture Division", |
| "model": "manara-annotate-v1", |
| "declaration_date": "2026-02-23" |
| }, |
|
|
| "data_residency": { |
| "policy": "100% Local Inference", |
| "description": "All inference, training, and data processing operations are designed to execute entirely within the operator's sovereign infrastructure. No data is transmitted to external, third-party, or cloud services during inference.", |
| "allowed_regions": ["Kuwait", "GCC Member States", "Designated Sovereign Cloud Zones"], |
| "cross_border_transfer": false, |
| "data_at_rest_encryption": "AES-256", |
| "data_in_transit_encryption": "TLS 1.3" |
| }, |
|
|
| "inference_sovereignty": { |
| "external_api_calls": false, |
| "telemetry_collection": false, |
| "model_phoning_home": false, |
| "offline_capable": true, |
| "edge_deployment_formats": ["GGUF (llama.cpp)", "ONNX", "TensorRT"], |
| "recommended_hardware": "NVIDIA A100 / H100 (on-premise) or sovereign cloud GPU instances" |
| }, |
|
|
| "regulatory_compliance": { |
| "kuwait_declaration_2026": { |
| "status": "Compliant", |
| "principles_met": [ |
| "Transparency & Explainability", |
| "Human Oversight & Control", |
| "Fairness & Non-Discrimination", |
| "Privacy & Data Protection", |
| "Security & Resilience", |
| "Accountability" |
| ] |
| }, |
| "kuwait_dppr": { |
| "status": "Compliant", |
| "pii_handling": "All PII identified, labeled, and redacted prior to training ingestion", |
| "data_subject_rights": "Supported via redaction audit pipeline" |
| }, |
| "kuwait_national_ai_strategy_2025_2028": { |
| "status": "Aligned", |
| "pillar": "Sovereign AI Infrastructure & Arabic-First Capabilities" |
| }, |
| "iso_iec_42001": { |
| "status": "Aligned", |
| "description": "AI Management System documentation provided" |
| } |
| }, |
|
|
| "model_governance": { |
| "human_oversight": true, |
| "automated_decision_making_without_review": false, |
| "bias_audit_completed": true, |
| "bias_mitigation_strategy": "Multi-dialect balanced dataset; Western-bias reduction via Arabic-only training corpus", |
| "sharia_compliance_filtering": true, |
| "explainability_method": "Confidence scoring on all label outputs", |
| "audit_trail": "Full data lineage documented in data_distillation/" |
| }, |
|
|
| "contact": { |
| "architect": "Lead Sovereign AI Architect", |
| "organization": "AgenThink", |
| "email": "farouq@agenthink.ai", |
| "linkedin": "farouq sultan" |
| } |
| } |
|
|
