Enforce Mode can operate as an inline control for connected AI traffic. Configured policy, redaction and evidence logging are applied before a request reaches an approved endpoint. Integration effort depends on the customer architecture and enforcement pattern.
Enforce Mode evaluates requests routed through configured control paths before provider execution. Policy, sensitive-data handling and evidence coverage depend on the selected architecture, integration and customer configuration.
Define governance rules for covered workloads and approved models. Configured policy can block, redact, route or escalate based on available content, identity, department and classification fields.
Automatically detect and redact personally identifiable information before it reaches any AI model. Supports Australian Privacy Act definitions, Medicare numbers, TFNs, and custom entity types.
Detect and block prompt injection attacks, jailbreak attempts, and adversarial inputs in real time. Protect your AI systems from manipulation without slowing legitimate workflows.
Route high-risk or sensitive requests to a human approver before they proceed. Configurable approval workflows with full audit trail — ideal for regulated industries and government use cases.
Configure routing for sensitive or classified data to approved endpoints. Residency outcomes depend on the selected deployment, model providers, integrations and customer controls.
Every request, every policy decision, every redaction — logged immutably with full context. Tamper-evident records ready for compliance audits, compliance assessments, and regulatory review.
Gateway policy evaluation adds less than 1ms to request latency. Enterprise-grade throughput with no perceptible impact on end-user experience — even at scale.
Deploy Gateway as a transparent proxy in front of your existing AI integrations. No SDK changes, no application refactoring — your teams keep working exactly as they do today.
Real-time dashboards showing policy trigger rates, blocked requests, redaction volumes, and approval queue status. Understand your AI risk posture at a glance.
Every AI request from every application in your organisation flows through Enforce Mode. Here's what happens in under a millisecond.
Your application sends an AI request as normal. Gateway intercepts it transparently — no SDK changes, no proxy configuration visible to developers. The request is captured with full metadata: user identity, department, application, timestamp, and data classification.
The request is evaluated against your active policy set in real time. Policies can inspect content, user attributes, data classification, model selection, and request context. Evaluation is complete in under 1ms — imperceptible to end users.
Sensitive data is detected using Australian-specific entity recognition — Medicare numbers, TFNs, ABNs, health identifiers, and custom types. PII is redacted or tokenised before the request proceeds, with the original preserved in the secure audit log.
Based on policy outcome, the request is routed to the appropriate model (including sovereign-only routing for sensitive data), escalated to a human approver, or blocked with a structured response returned to the application. All outcomes are logged.
Covered requests, policy decisions, redactions and responses can be written to a tamper-evident governance log. Evidence coverage and retention depend on source completeness, configuration and privileged administration and can support customer review.
Start with Observe Mode for immediate visibility. Expand to Enforce Mode for inline enforcement. Complete the journey with Orchestrate Mode for full sovereign agentic governance.
Our structured POC puts Gateway in front of your AI traffic in Week 1 and delivers a full executive readout in Week 2 — with zero risk and zero infrastructure changes.
Explore the live platform — real dashboards, model routing, policy engine, and audit logs. No sign-up required.
