Glassbox · compliance
Alberta AI Usage Policy alignment
Line-by-line correspondence between Alberta's public-sector AI policy and the Glassbox implementation. Glassbox is not endorsed by the Ministry; this page documents our design choices, not an external audit.
| Policy requirement | Glassbox implementation | Status |
|---|---|---|
| Sovereign compute | Self-hosted federal and Alberta provincial data via the Render replica. No third-party AI involved in pattern detection. Calibrated-language enforcement is local regex; semantic retrieval is optional with explicit opt-in. | Met |
| Open-source models | All detection logic open-source under MIT. Pattern detectors live in src/lib/patterns/ as inspectable, testable modules. | Met |
| Audit trails | Every output carries an audit token chained to source-data tokens. The 'why this match?' trail can always be walked back to the source row. | Met |
| Accountability for outputs | Calibrated-language discipline (no fraud claims, no causal language); citation rigor (every match cites at least one source row from a named dataset). | Met |
| Cybersecurity gates | Deployment-stage requirement. AWS infrastructure scaffolding lives in infra/aws/security/ with placeholder Terraform stubs; production gates ship before any non-development deploy. | Deployment-stage |