Data Privacy Gate Check Prompt Template

This prompt is designed for privacy engineers, DPOs, and compliance engineering teams who need to evaluate a proposed AI processing action against a specific data protection policy before execution. It acts as an automated gate that produces a structured decision with a data protection impact summary, purpose limitation check, and consent verification status. Use this prompt when you need a reviewable, evidence-grounded assessment that blocks non-compliant actions and documents the reasoning. The ideal workflow involves piping a proposed action description and the relevant policy text into the prompt and receiving a machine-readable pass, fail, or needs-review decision that your application can enforce.

This prompt is not a replacement for legal counsel or a full Data Protection Impact Assessment (DPIA). It is an operational control for high-volume or automated AI workflows where manual review of every action is impractical. Do not use this prompt when the processing action involves special category data, high-risk profiling, or novel processing purposes that have not been previously reviewed by a human. The prompt requires the exact policy text as grounding context; without it, the model will improvise and produce unreliable results. Always configure your harness to block execution when the policy text is missing or when the model returns a needs-review or fail decision.

Before deploying this prompt, define the exact actions that constitute processing in your system, map them to the policy sections that govern them, and test the prompt against a golden set of known compliant and non-compliant scenarios. The prompt's value comes from its consistency and auditability, not from nuanced legal interpretation. Wire the output into your application's action gating logic so that a fail or needs-review decision prevents execution and logs the full reasoning for later review. If your use case involves multiple jurisdictions or overlapping policies, run separate gate checks for each applicable policy rather than combining them into a single prompt.

The following template is a complete, copy-ready prompt for a Data Privacy Gate Check. It is designed to be used as a system prompt, with the specific action to be evaluated passed as the user message. The prompt instructs the model to act as a privacy engineer, grounding its assessment strictly in the provided policy text and producing a structured JSON output. Before using this template, you must prepare your specific data handling policy text and define the risk tolerance for your use case.

markdownCopy
You are a privacy engineering analyst. Your task is to perform a Data Protection Impact pre-screening on a proposed AI processing action. You must ground your entire assessment in the provided [POLICY_TEXT]. Do not use external knowledge. If the policy does not address a point, state that it is not covered.

Evaluate the proposed action against the following criteria from the policy:
1.  **Purpose Limitation**: Is the processing purpose compatible with the original collection purpose?
2.  **Data Minimization**: Is the data requested adequate, relevant, and limited to what is necessary?
3.  **Consent & Legal Basis**: Is there a valid legal basis and, if consent is required, has it been verified?
4.  **Data Residency**: Does the processing location comply with data residency requirements?
5.  **Sensitive Data**: Does the action involve special categories of data, and are the required safeguards in place?

**Inputs:**
- **Proposed Action**: [ACTION_DESCRIPTION]
- **Data Categories**: [DATA_CATEGORIES]
- **Processing Region**: [PROCESSING_REGION]
- **User Consent Status**: [CONSENT_STATUS]
- **Policy Text**: [POLICY_TEXT]

**Output Requirements**:
You must produce a valid JSON object with the following schema. Do not include any text outside the JSON block.
{
  "gate_decision": "PASS" | "FAIL" | "NEEDS_REVIEW",
  "overall_risk_level": "LOW" | "MEDIUM" | "HIGH" | "CRITICAL",
  "checks": [
    {
      "check_name": "string",
      "status": "PASS" | "FAIL" | "REVIEW_REQUIRED" | "NOT_COVERED",
      "policy_reference": "string or null",
      "rationale": "string"
    }
  ],
  "blocking_conditions": ["string"],
  "recommended_actions": ["string"]
}

**Constraints**:
- If the [RISK_LEVEL] is set to "STRICT", any single "FAIL" status must result in an overall "FAIL" gate decision.
- If the [CONSENT_STATUS] is "NOT_VERIFIED" and the legal basis is consent, the consent check must "FAIL".
- Cite specific clauses from the [POLICY_TEXT] in the `policy_reference` field whenever possible.

To adapt this template, replace the square-bracket placeholders with your specific inputs. The [POLICY_TEXT] placeholder is critical; it should contain the full text of your internal data handling policy, GDPR article, or CCPA section. The [RISK_LEVEL] variable allows you to toggle between a lenient mode that flags issues for review and a strict mode that automatically blocks any non-compliant action. After execution, always validate the output JSON against the defined schema before allowing the gate decision to control any downstream workflow. For high-risk actions, the output should be routed to a human reviewer regardless of the model's gate_decision.

The Data Privacy Gate Check prompt is not a standalone chatbot interaction; it is a deterministic gate in a processing pipeline. It should be wired into an application as a pre-action check that runs before any AI system processes, stores, or shares personal data. The application must provide the prompt with the exact policy text to ground the decision, the proposed processing action, the data categories involved, the purpose of processing, and the user's consent record. Without these inputs, the gate cannot produce a reliable decision and should default to a blocking state.

The implementation harness should treat the model's output as a structured JSON object with fields for gate_decision (allowed, blocked, needs_review), purpose_limitation_check, consent_verification_status, and data_protection_impact_summary. Validate this output against a strict schema before acting on it. If the model fails to return valid JSON after a configurable number of retries (we recommend two), the harness must escalate to a human review queue and block the action. Log every gate decision, including the full prompt context, the model's raw output, the validation result, and the final action taken, to create an auditable record. For high-risk processing actions, such as those involving special category data or cross-border transfers, always require human approval regardless of the model's gate_decision.

Choose a model with strong instruction-following and JSON mode capabilities for this workflow. The prompt is designed for models that support structured output constraints. If you are using a model without native JSON mode, add a post-processing step to extract and repair the JSON, but flag any repaired output for human review. Do not use this prompt with models that are prone to hallucinating legal citations or policy interpretations without grounding. The provided policy text is the only source of truth; the model must not supplement it with external knowledge. Implement a source-grounding check in your evaluation suite that verifies every cited policy reference in the output exists in the input policy text. Finally, never allow a gate_decision of allowed to bypass application-level enforcement; the gate is an advisory control that must be paired with hard-coded policy rules and human oversight for regulated data.