brando:Policy

Governance node encoding guard rails, refusals, compliance tags, and retrieval visibility.

rdfs:comment
Governance node encoding guard rails, refusals, compliance tags, and retrieval visibility.

---

Term definition

• IRI: brando:Policy
• Type: owl:Class
• Superclass: schema:Intangible
• Defined in: brando: (Brando Schema Vocabulary v1.3)

brando:Policy is the core governance type in Brando. It represents a machine-readable policy object that AI runtimes and Brand OS components can:

• load,
• apply,
• and log against,

for a given brand, context, product, or category.

Typical policies cover:

• brand safety and tone control,
• regulatory constraints,
• refusal behaviour,
• retrieval/visibility rules for AI systems.

---

Intended usage

Use brando:Policy whenever you need a governed, inspectable rule set that constrains brand behaviour in AI systems.

Common patterns:

• Brand-level “global” policies
  • e.g. generic brand safety, tone rules, and prohibited terms.
• Context-level policies
  • e.g. support chatbot policies, investor relations policies, region-specific restrictions.
• Product- or category-specific policies
  • e.g. rules for high-risk product lines, GTIN-specific constraints.

In a typical flow:

1. Your runtime resolves a brando:Brand and brando:Context.
2. It collects all brando:Policy instances linked via brando:hasPolicy and/or brando:appliesTo*.
3. It uses policy properties (e.g. brando:guardRails, brando:mustNotDos, brando:refusalStrategies, brando:enforcementLevel) to:
   • construct prompts,
   • enforce pre/post constraints,
   • drive reporting and audit.

---

Relationships

Inbound

Policies are usually reached from:

• brando:hasPolicy
• Domain: brando:Brand, brando:Context
• Range: brando:Policy
• Comment (summary): Attaches governance, safety, and compliance policies to a brand or context.

Example:

{
  "@id": "https://example.com/brand/northstar",
  "@type": "brando:Brand",
  "schema:name": "Northstar Bank",
  "brando:hasPolicy": [
    { "@id": "https://example.com/brand/northstar/policies/global-brand-safety" },
    { "@id": "https://example.com/brand/northstar/policies/regulated-claims-uk" }
  ]
}

Policies can also be scoped via product- and GTIN-level links:

• brando:appliesToProduct (range: gs1:Product)
• brando:appliesToGTIN (range: xsd:string)

These are properties whose domain includes brando:Policy, allowing fine-grained scoping.

---

Key properties (domain includes brando:Policy)

The following properties have brando:Policy in their rdfs:domain. Several are also used on other classes (e.g. brando:Context, brando:AutomationRule); here they are described specifically from the policy perspective.

1. Scope and applicability

Property IRI	Range	Type	Comment (from vocab, summarised)
brando:appliesTo	xsd:string	owl:DatatypeProperty	Optional scoping for which products, categories, personas, or channels this policy applies to.
brando:appliesToGTIN	xsd:string	owl:DatatypeProperty	GTIN(s) for which this branded category or policy is applicable.
brando:appliesToProduct	gs1:Product	owl:ObjectProperty	Links a branded category or policy to one or more GS1 Product nodes.
brando:effectiveDuring	xsd:string	owl:DatatypeProperty	Time window or schedule when this context, campaign, or policy is active, expressed as structured JSON.

Usage notes (non-normative):

• brando:appliesTo is deliberately free-form (string). You can encode:

  • simple labels: "paid social; EMEA",
  • or structured JSON as a string: {"personas":["Everyday saver"],"channels":["chatbot"]}.

• brando:appliesToGTIN is useful for granular, product-level constraints in commerce use cases.

• brando:effectiveDuring is intended for JSON-as-string, e.g.:

```json { "validFrom": "2025-01-01", "validThrough": "2025-12-31" }

```

---

2. Guard rails, refusals, and risk

Property IRI	Range	Type	Comment (from vocab, summarised)
brando:guardRails	xsd:string	owl:DatatypeProperty	Non-negotiable boundaries that AI systems must not cross.
brando:refusalStrategies	xsd:string	owl:DatatypeProperty	Preferred patterns for declining or redirecting unsafe or out-of-scope requests.
brando:riskScenarios	xsd:string	owl:DatatypeProperty	Known edge cases and misuse patterns requiring caution or escalation.
brando:complianceTags	xsd:string	owl:DatatypeProperty	Regulatory or legal tags such as “GDPR”, “HIPAA”, or “FCA”.
brando:riskTag	xsd:string	owl:DatatypeProperty	Lightweight tags describing primary risk domains (e.g. brand-safety) for querying, reporting, and dashboards.

Usage notes (non-normative):

• Treat guardRails, refusalStrategies, and riskScenarios as lists of strings (one per bullet), either:

  • modelled as a JSON array string, or
  • represented as repeated property values in JSON-LD.

• Use complianceTags and riskTag as query-friendly labels to filter and report policies.

Example pattern:

{
  "@id": "https://example.com/brand/northstar/policies/regulated-claims-uk",
  "@type": "brando:Policy",
  "brando:guardRails": [
    "Do not imply guaranteed returns.",
    "Do not provide personalised investment advice."
  ],
  "brando:refusalStrategies": [
    "Politely explain that you cannot provide personalised investment advice.",
    "Offer to connect the user with a human advisor."
  ],
  "brando:riskScenarios": [
    "User asks if a specific product is 'risk-free'.",
    "User requests a guaranteed level of return."
  ],
  "brando:complianceTags": [
    "FCA",
    "UK-financial-promotions"
  ],
  "brando:riskTag": [
    "financial-regulation",
    "consumer-protection"
  ]
}

---

3. Enforcement, visibility, and retrieval

Property IRI	Range	Type	Comment (from vocab, summarised)
brando:enforcementLevel	xsd:string	owl:DatatypeProperty	Indicates how strongly a policy or automation rule must be applied (suggested values include “mandatory”, “advisory”, “conditional”).
brando:priority	xsd:string	owl:DatatypeProperty	Governance precedence for this policy when multiple rules could apply.
brando:visibilityRating	xsd:string	owl:DatatypeProperty	Qualitative indicator of how easily this policy should be discoverable by AI systems.
brando:retrievableInLLM	xsd:boolean	owl:DatatypeProperty	Boolean flag indicating whether this policy may be indexed or retrieved by language models.
brando:embeddingQuality	xsd:string	owl:DatatypeProperty	Qualitative or scored assessment of how well this content performs in vector search or RAG.

Usage notes (non-normative):

• enforcementLevel is shared with brando:AutomationRule. Common string values (not enforced by the vocab) include:

  • "mandatory",
  • "advisory",
  • "conditional".

• priority can be:

  • ordinal (e.g. "1-high", "2-medium", "3-low"), or
  • descriptive (e.g. "portfolio-level", "brand-level", "context-level").

• Use visibilityRating, retrievableInLLM, and embeddingQuality to tune how and where policies appear in:

  • RAG results,
  • AI-readable policy searches,
  • agentic tooling.

Example:

{
  "@id": "https://example.com/brand/northstar/policies/global-brand-safety",
  "@type": "brando:Policy",
  "brando:enforcementLevel": "mandatory",
  "brando:priority": "1-high",
  "brando:visibilityRating": "high",
  "brando:retrievableInLLM": true,
  "brando:embeddingQuality": "good"
}

---

4. Lifecycle and process

Property IRI	Range	Type	Comment (from vocab, summarised)
brando:reviewWorkflow	xsd:string	owl:DatatypeProperty	Description of human and/or automated steps for reviewing content and policy changes.
brando:updatePolicy	xsd:string	owl:DatatypeProperty	Rules for how often, by whom, and under what process policies are updated.

Usage notes (non-normative):

• These properties describe the governance process – they don’t implement it.
• Combine them with temporal properties in instance data (dcterms:created, dcterms:modified) for full lifecycle tracking.

Example:

{
  "@id": "https://example.com/brand/northstar/policies/regulated-claims-uk",
  "@type": "brando:Policy",
  "brando:reviewWorkflow": "Compliance team reviews changes weekly. Any exceptions must be approved by legal.",
  "brando:updatePolicy": "Review at least quarterly or when regulations change."
}

---

5. Files and integrations

These properties are shared across several Brando types but are frequently used on policies:

Property IRI	Range	Type	Comment (from vocab, summarised)
brando:downloadableFile	schema:MediaObject	owl:ObjectProperty	Links a brand, token, context, campaign, policy, or category to the most up-to-date creative or reference files exposed by the Brand OS or DAM.
brando:supportingFile	schema:MediaObject	owl:ObjectProperty	Links a node to additional files that are helpful for production but are not the primary source-of-truth asset.
brando:externalSystem	schema:SoftwareApplication	owl:ObjectProperty	Links a brand, context, policy, campaign, or automation rule to a software application that represents an integration target.
brando:externalSystemId	xsd:string	owl:DatatypeProperty	Identifier for the corresponding object in an external system such as Adobe GenStudio, Meta Ads Manager, Figma, or internal CMSs.

Usage notes (non-normative):

• For policy nodes:

  • downloadableFile might point to full legal policy PDFs or structured policy docs.
  • supportingFile might point to training decks, explainer docs, or example cases.

• externalSystem + externalSystemId can be used to:

  • map a policy to a specific enforcement configuration in another platform,
  • or drive sync jobs between the Brand Knowledge Graph and external governance tools.

Example:

{
  "@id": "https://example.com/brand/northstar/policies/regulated-claims-uk",
  "@type": "brando:Policy",
  "brando:downloadableFile": {
    "@id": "https://assets.example.com/northstar/policies/regulated-claims-uk.pdf",
    "@type": "schema:MediaObject",
    "schema:name": "Regulated Claims Policy – UK (PDF)"
  },
  "brando:externalSystem": {
    "@id": "https://example.com/integrations/policy-engine"
  },
  "brando:externalSystemId": "northstar_regclaims_uk_v2"
}

---

Example: Brand + Context + Policies

A combined example showing how brando:Policy is used with brando:Brand and brando:Context:

{
  "@context": {
    "schema": "https://schema.org/",
    "brando": "https://brandoschema.com/",
    "gs1": "https://ref.gs1.org/voc/"
  },
  "@graph": [
    {
      "@id": "https://example.com/brand/northstar",
      "@type": "brando:Brand",
      "schema:name": "Northstar Bank",
      "brando:hasContext": {
        "@id": "https://example.com/brand/northstar/context/retail-chatbot"
      },
      "brando:hasPolicy": {
        "@id": "https://example.com/brand/northstar/policies/global-brand-safety"
      }
    },
    {
      "@id": "https://example.com/brand/northstar/context/retail-chatbot",
      "@type": "brando:Context",
      "brando:audienceSegment": "Retail banking customers",
      "brando:domainContext": {
        "channel": "chatbot",
        "surface": "public-website",
        "region": "UK"
      },
      "brando:hasPolicy": {
        "@id": "https://example.com/brand/northstar/policies/regulated-claims-uk"
      }
    },
    {
      "@id": "https://example.com/brand/northstar/policies/global-brand-safety",
      "@type": "brando:Policy",
      "brando:enforcementLevel": "mandatory",
      "brando:priority": "1-high",
      "brando:guardRails": [
        "Do not output hate speech, harassment, or discriminatory language.",
        "Do not claim features or capabilities the brand does not provide."
      ],
      "brando:riskTag": [
        "brand-safety",
        "tone-of-voice"
      ],
      "brando:retrievableInLLM": true
    },
    {
      "@id": "https://example.com/brand/northstar/policies/regulated-claims-uk",
      "@type": "brando:Policy",
      "brando:enforcementLevel": "mandatory",
      "brando:appliesTo": "Retail banking products in the UK",
      "brando:appliesToGTIN": "12345678901234",
      "brando:appliesToProduct": {
        "@id": "https://example.com/gs1/product/12345678901234",
        "@type": "gs1:Product"
      },
      "brando:guardRails": [
        "Do not promise guaranteed investment returns.",
        "Do not position speculative products as suitable for all customers."
      ],
      "brando:refusalStrategies": [
        "Explain that you cannot provide personalised investment advice.",
        "Encourage the user to speak to a regulated advisor."
      ],
      "brando:complianceTags": [
        "FCA",
        "UK-financial-promotions"
      ],
      "brando:riskTag": [
        "financial-regulation",
        "consumer-protection"
      ],
      "brando:reviewWorkflow": "Compliance reviews this policy monthly and after any regulatory change.",
      "brando:updatePolicy": "Update within 30 days of new FCA guidance affecting retail financial promotions.",
      "brando:retrievableInLLM": true
    }
  ]
}

Runtime pattern (non-normative):

1. Resolve the current brando:Brand and brando:Context.

2. Collect:

   • global policies attached to the brand,
   • context-specific policies (e.g. regulated-claims-uk),
   • product-level policies via appliesToProduct / appliesToGTIN, if a product is in scope.

3. Merge policy fields (guardRails, mustNotDos, refusalStrategies, etc.) according to:

   • enforcementLevel,
   • priority,
   • your own precedence rules.

4. Use this merged configuration to:

   • build prompts,
   • enforce pre/post checks,
   • log which policies were in effect for each interaction.

---

Related documentation

• Concepts / overview

  • Brando types overview
  • Getting started

• Related types

  • brando:Brand
  • brando:Context
  • brando:AutomationRule
  • brando:BrandedCategory

• Properties index

  • Properties overview