Long-form, versioned reference documents, guides, and research on deploying AI safely in production environments. Not blog posts. Citable, maintained, authoritative.
Deep reference guides for all eight Production Safety Framework domains. Aligned with AIDA, CAIS, CPAP, and CPAA examination content. Licensed CC BY 4.0.
Prompt injection defence, input validation, intent classification, and sanitisation practices.
Schema enforcement, content filtering, hallucination detection, and confidence gating.
PII handling, consent chains, data residency, vector store risk, and erasure obligations.
Inference logging, quality scoring, drift detection, and alerting architecture.
Canary releases, rollback procedures, circuit breakers, and fallback path design.
Autonomy levels, escalation design, blind review sampling, and skill maintenance.
Threat modelling, credential management, adversarial robustness, and red-teaming.
Abstraction layers, version pinning, fallback providers, and deprecation response planning.
The following reference documents are currently being written and will be published over the coming weeks.