Platform Platform
System
ConceptsEnginePolicy as codeDeclarationsSafe changeGatewaysIntegrationsObservabilityAdministrationSecurityHuman reviewAudit and evidenceData retentionSecrets and data classification
Controls
Registries and documentationAuthentication and authorizationInjection detectionData redactionCode fingerprintingRole and judge checksContent classificationSpend and loop limitsBusiness rules
Solutions Solutions
By what you do
Sell into the enterpriseControl the AI you run
By industry
Financial servicesDigital assetsInsuranceHealthcareLegalUser-generated content
By discipline
AI governanceTrust and safetyRisk and compliance
Cases Cases Embedded control planeSource-code leakTrading agents over MCPLive firehoseRefund assistant
Compare Compare LiteLLMNVIDIA NeMo GuardrailsOPAROOSTAgent Governance Toolkit
Resources Resources
Guides
Enterprise review questionsPrompt injectionAgent and control layerAgent architecturesDecision system mapAI control maturity model
Standards
Standards OWASP Agent Control StandardEU AI ActPMI AI standardNIST AI RMFERC-8004
Book a demo
Platform · System

As code means the lifecycle. Not the language.

Versions, review, promotion, rollback, an audit trail. Borrowed from software, which solved this decades ago. It does not mean someone has to write a program to change a threshold.

Rules only, so changing one needs no release.

Read this and tell us what it does

YAML, versioned, backtestable
first_time_destination_review:
  all:
    - path: "event.type"
      op: eq
      value: "transfer_intent"
    - path: "event.data.destination_known"
      op: eq
      value: false
    - path: "event.data.amount_usd"
      op: gte
      value: "{{ constants.unknown_destination_amount_floor }}"
  effects:
    verdict: flagged
    response:
      reason: "First-time destination above threshold"
    actions:
      - action: unknown_destination_case

You just did, and you are not an engineer on this project. That is the entire argument. Conditions on the event, a verdict, a reason the user will see, and the case it opens.

If you have looked at a policy engine before, you have seen the other kind: policy is a programming language, and every threshold change is a ticket. That is why your rules are out of date: the person who knows what the rule should say cannot write it, and the person who can is working on something else.

Who is allowed to do what

RightTypically held by
Edit a draftthe policy owner: risk, compliance, trust and safety
Move it to candidate and measure itthe same person
Promote it to productiona separate right, granted separately

Every edit is recorded against the person who made it.

Still readable to your engineers

It is a text file. It lives in your repository if you want it to, gets reviewed in a pull request if that is how you work, and diffs like anything else.

Related: how a change reaches production · what a stateful rule can read
Book a demo