ThirdLaw AI Control Layer

ThirdLaw AI Control Layer is a runtime enforcement platform for AI applications and agents. It captures AI interactions, evaluates them against enterprise-defined policies, intervenes in real time, and supports post-incident investigation. Core Workflow: - Collect: Captures prompts, outputs, tool calls, and agent actions from AI sessions. Ingests data via API gateways (Envoy, Kong, NGINX), SDKs (Python, TypeScript, Go), OpenTelemetry instrumentation, or direct API pipelines. - Evaluate: Assesses AI behavior against enterprise policies using semantic checks, classifiers, LLM-based reasoning, and rules-based logic. Detects policy violations and anomalous patterns across sessions and agent chains. - Intervene: Applies runtime enforcement actions including blocking, redacting, rerouting, or escalating interactions that violate policy. Policies can be scoped by app, route, role, user group, model, or tool. - Investigate: Provides session replay, event search, and tool-call forensics for incident review. Routes findings and context to SIEM, SOAR, and ITSM systems to support compliance and response workflows. Policy Management (called "Laws"): Policies are authored in plain language, translated into runtime logic, validated on example traffic, versioned, and published with scoped rollout controls. Deployment Options: - Managed single-tenant SaaS (isolated AWS account and VPC) - Customer-hosted AWS VPC - Self-managed Kubernetes via Helm (on-prem or private cloud) The platform targets security and IT teams managing AI application risk and is aligned with the Gartner AI TRiSM category.