Compliance Platform¶
The Geodesia G-1 Compliance Platform provides the complete toolchain to operate, audit, and govern an AI system in compliance with the EU AI Act and 12 other global regulatory frameworks. It runs on G-1 Studio (/v1/glad/*) and is exposed both as an API and through the built-in web interface. The complete endpoint list is in the Studio API map; which UI control calls what is in the UI Component Reference.
Call it¶
The one call that tells you whether the deployment is defensible right now.
overall_health is green / yellow / red, and issues / warnings are already written for a human to read. Details: Dashboard.
Platform Components¶
| Component | API prefix | Description |
|---|---|---|
| Dashboard | /v1/glad/dashboard | Real-time operational metrics: call volume, passed/blocked/flagged counts, safety and hallucination rates |
| FRIA | /v1/glad/fria | Fundamental Rights Impact Assessment — EU AI Act Article 27 dossier creation, management, and PDF/DOCX export |
| Human Oversight | /v1/glad/oversight | Queue of flagged calls requiring human review; tiered escalation (Operator → AI Responsible); decision recording |
| Kill Switch | /v1/glad/kill-switch | Instant service suspension; compliance-aware deactivation within configured time windows |
| Audit Chain | /v1/glad/chain/status, /v1/glad/chain/verify | HMAC-linked append-only log; cryptographic integrity verification |
| Watermark | /v1/glad/watermark/verify | HMAC-SHA256 latent AI watermark; verification endpoint |
| Reports | /v1/glad/report, /v1/glad/deployer-manual | PDF/DOCX audit bundles and deployer transparency manuals |
| Models | /v1/glad/models/available, /v1/glad/models/switch | Available checkpoint catalog; model switching |
| Threshold Prefs | /v1/glad/threshold-prefs | Deployer-specific detection thresholds stored in the database |
| Retention | /v1/glad/retention/status | Data retention policy state and what is past its window |
| Provider Identity | /v1/glad/provider-identity | Machine-readable provider identity (for AI Act Article 13) |
| License Tokens | /v1/glad/license-tokens | Customer license token management |
Compliance Architecture¶
Every inference call goes through the following compliance pipeline regardless of whether you use the gateway or the evaluate endpoint:
Every inference call passes through the same compliance pipeline — whether it arrives via the gateway or the evaluate endpoint.
All data is written to a single SQLite database (var/glad.sqlite3 by default) that is shared between the gateway and the product backend.
Supported Regulatory Frameworks¶
EU_AI_ACT GDPR ISO_42001 NIST_AI_RMF CA_SB_942 ITALY_132_2025 UK_DUAA_2025 BRAZIL_2338 CANADA_AIDA CHINA_GB45654 COLORADO_SB21_169 NYC_LL144 SOC2
See Regulatory Coverage for detailed mapping.
Quick Reference: Common Compliance Tasks¶
| Task | Endpoint | Method |
|---|---|---|
| Check if the service is compliant | GET /v1/glad/scorecard | GET |
| Create a FRIA dossier | POST /v1/glad/fria | POST |
| Export a FRIA as PDF | GET /v1/glad/fria/{id}/export?fmt=pdf | GET |
| View pending human reviews | GET /v1/glad/oversight/pending | GET |
| Record a review decision | POST /v1/glad/oversight/decide | POST |
| Activate the kill switch | POST /v1/glad/kill-switch/activate | POST |
| Verify audit chain integrity | GET /v1/glad/chain/verify | GET |
| Verify a watermark | POST /v1/glad/watermark/verify | POST |
| Generate a compliance report | POST /v1/glad/report | POST |
| Generate a deployer manual | POST /v1/glad/deployer-manual | POST |
| Set detection thresholds | POST /v1/glad/threshold-prefs | POST |
| List available model checkpoints | GET /v1/glad/models/available | GET |