Sensitive Document Redaction Check
Create a precise redaction plan for a document or screenshot before external sharing while preserving the information needed for the task.
Users preparing screenshots, PDFs, contracts, logs, support tickets, invoices, research material, or internal documents for public posts, AI tools, clients, or collaborators.
Act as a careful redaction planner. Review the described document, screenshot, transcript, log, or form before it is shared outside its original context. Do not repeat any sensitive values that appear in the input.
Document type and purpose:
{{document_type_and_purpose}}
Recipient or publication channel:
{{recipient_or_channel}}
Visible fields or redacted sample:
{{visible_fields_or_sample}}
Create the review using this structure:
1. Sharing objective: identify the exact information the recipient needs and what is unrelated.
2. Direct identifiers: flag names, faces, signatures, email addresses, phone numbers, street addresses, account numbers, IDs, barcodes, QR codes, usernames, and unique case references.
3. Sensitive content: flag financial, health, employment, legal, academic, customer, child-related, security, credential, private-message, and confidential business information.
4. Hidden and indirect exposure: check timestamps, locations, filenames, browser tabs, notifications, comments, tracked changes, revision history, document properties, embedded links, image metadata, background windows, and combinations that can re-identify someone.
5. Redaction plan: label each item Keep, Remove, Mask, Generalize, Crop, Replace with synthetic data, or Obtain permission. Explain the minimum necessary transformation.
6. Technical verification: explain how to confirm that redaction is irreversible, including flattening or exporting, checking searchable text layers, removing metadata, and reopening the final file. Do not assume drawing a black box permanently removes underlying content.
7. Final checklist: provide a short pre-send inspection for both content and destination.
8. Decision: Safe after listed changes, Needs authorized review, or Do not share through this channel.
Rules: Never request the unredacted value of a password, token, private key, full financial account number, or government ID. Do not claim that a document is legally anonymized. Recommend specialist review for regulated records, legal evidence, children, patients, employees, or large-scale disclosures.How to use
- Describe the document type, recipient, purpose, and every visible field or layer without exposing secret values.
- Apply changes to a separate copy and preserve the original only in its authorized location.
- Verify the final export by reopening it, searching or selecting text, and checking metadata and hidden layers.
- Use authorized privacy or legal review for regulated, evidentiary, employee, patient, or child-related records.
Example input
Document: a screenshot of a support dashboard for a public bug report. Visible fields: customer name, email, ticket ID, message, browser address bar, two Slack notifications, account-plan label, and current time. Purpose: show the UI error to a software vendor.
Example output
Keep the error message and relevant UI state. Remove or mask the customer name, email, ticket ID, account-plan label, and any unique internal identifiers. Crop the browser address bar if it contains a private tenant URL. Hide Slack notifications and unrelated background windows. Generalize the time unless timing is necessary to reproduce the bug. Export a flattened copy, reopen it, test text selection, and inspect metadata before sending.
Customization tips
- — Ask for a page-by-page plan for long PDFs or a column-by-column plan for exported tables.
- — Include the target format because screenshots, PDFs, word-processing files, and spreadsheets hide data differently.
- — State whether the recipient needs to identify individuals or only understand a pattern or issue.
Tags
FAQ
Is covering text with a black rectangle enough?
What is indirect identification?
Should metadata be removed?
Can this prompt legally anonymize a dataset or document?
Why create a separate copy?
Related Prompts
All Safety & Review prompts →Privacy-Safe Data Sharing Review
Review text, files, screenshots, or datasets before sharing them with an AI tool, vendor, colleague, or public audience.
Vendor Security & Compliance Risk Review
Systematically evaluate a third-party vendor or tool for security, data privacy, and compliance risk before integrating it into your business.
AI Output Fact-Check & Evidence Audit
Audit an AI-generated answer for unsupported claims, weak citations, invented details, outdated facts, and misleading certainty before reuse.
Related Skills
Privacy and Data Handling Reviewer
Review a product or workflow for privacy and data-handling risks, then create practical controls, documentation needs, and remediation priorities.
AI Governance & Compliance Checklist
Reviews an AI-related workflow, tool, or output against a practical governance checklist covering data privacy, bias, accountability, and transparency.
Sensitive Data Redaction Workflow Designer
Design redaction workflows for documents, logs, prompts, exports, and datasets using detection, review, masking, validation, access, and evidence.
Related Articles
AI Privacy Checklist Before Launching an AI App
A practical, non-legal checklist covering data collection, uploads, logging, admin access, and disclosure before launching an AI app.
How do I redact sensitive information before using ChatGPT or Claude?
Redaction means removing or masking personal, financial, customer, credential, health, and confidential business details before sharing text with AI.