Smith & Gray
SolutionsIndustriesServicesPlatformPricingBlog
Free AI AuditAbout
All platform capabilitiesPlatform Capability

Guardrails & Safety Engine

Every message — in and out — passes through a multi-layer safety engine. Pattern detection, personal information redaction, ensemble safety classification, and data sovereignty routing. No agent goes live without it.

The safety engine is the core of every Smith & Gray agent. Before the model sees a message, it is filtered for injection attempts and personal information is redacted. After the model responds, the output is classified by an ensemble of safety classifiers. If anything fails the confidence threshold, the agent blocks, logs to the immutable audit trail, and escalates to a human. Data sovereignty routing ensures sensitive data never leaves your jurisdiction. This is government-grade safety, built into every conversation.

What it includes

  • Multi-layer input filtering — injection detection before the model sees the message
  • Personal information redaction — sensitive data stripped before processing
  • Ensemble safety classification — multiple models check every response
  • Data sovereignty routing — sensitive data stays in your jurisdiction
  • Immutable audit trail — every decision logged, tamper-proof
  • Automatic human escalation — low-confidence actions flagged for review
  • Per-visitor abuse prevention — rate limiting and pattern detection

Safety controls

How this keeps AI safe

  • Fail-safe blocking — when in doubt, the agent blocks rather than risks harm
  • Australian Government ISM and Privacy Principles alignment
  • Every safety decision is logged with full context for audit
  • Confidence thresholds tuned per data classification (public, sensitive, restricted)

Which agents use this

Lead Agent — blocks prompt injection attempts and PII leaksCompliance Agent — routes sensitive data to approved jurisdictions onlyMarketing Agent — prevents brand-unsafe content from publishing

Want to see this in your infrastructure?

Book a free AI audit. We'll assess your infrastructure and recommend the right deployment approach.

Free AI Opportunity Call
Free AI Opportunity Call

Ask Smith

AI Assistant

Powered by Smith & Gray AI · Government-grade safety