AI Redaction

Last updated: 2026-03-23

AI-assisted redaction lets you instruct the AI agent to find and flag sensitive information across your case files. The agent searches for the content you describe, then opens a review panel where you verify and approve each match before any redactions are applied.

Instructing the AI agent for redaction

Open the AI agent panel and describe what you want to redact in plain language. Use specific instructions like "Redact all Social Security numbers," "Find and redact Jane Doe's phone number," or "Redact all email addresses in the case files." The agent understands categories of personally identifiable information as well as specific names, terms, and patterns.

The agent handles broad requests like "Redact all PII" as well as targeted ones like "Redact the account number on page 3 of Exhibit B." This makes AI-assisted redaction useful for both case-wide tasks and one-off corrections.

How AI redaction searches for sensitive content

After receiving your instruction, the agent searches case files for matching content. It examines both born-digital text and text detected in scanned documents, identifying matches by pattern (the format of a Social Security number) and by context (recognizing that a nine-digit number following "SSN:" is likely a Social Security number).

The search may take a moment depending on how many files are in your case and their size. The agent provides status updates in the chat as it processes documents.

The AI redaction review panel

When the agent finishes, it opens a review panel showing all detected matches. Each match displays the file name, page number, detected text, and surrounding context. A confidence indicator helps you quickly spot items that may need closer inspection.

Human review is required before any redactions are applied. The AI agent never redacts content on its own. You review the matches, select which ones to redact, and explicitly approve them. The final decision always rests with you.

Best use cases for AI redaction

AI-assisted redaction works best for these tasks:

  • Systematic PII removal - Redacting the categories of personal identifiers required by FRCP Rule 5.2 (Social Security numbers, taxpayer IDs, birth dates, minor names, financial account numbers) as well as phone numbers and email addresses across large document sets.
  • Named-entity redaction - Removing all references to a specific person, company, or account across multiple documents.
  • Pattern-based detection - Finding content that follows a specific format, such as account numbers, case reference numbers, or medical record numbers.
  • Large-scale review - When you have hundreds or thousands of documents that may contain the same types of sensitive information.

If you already know the exact terms to search for, the keyword search or batch redaction tools may be faster. AI-assisted redaction shines when you need the agent's ability to understand categories, patterns, and context.

Frequently asked questions

Does the AI agent redact content automatically?
No. The agent identifies and flags potential matches, but you must review and approve each one before any redactions are applied.
Can I give the agent complex instructions?
Yes. The agent understands natural language, so describe what to redact in as much detail as you need. For example, "Redact all phone numbers and email addresses belonging to minor children" is a valid instruction.
How accurate is the AI detection?
Each match includes a confidence score. High-confidence matches are generally accurate, while lower-confidence ones may need closer inspection. The review panel lets you check every match before approving it.
Can I use AI redaction on scanned documents?
Yes. The agent processes both born-digital text and text detected in scanned documents and images. It identifies sensitive information across the document formats Hintyr supports.

Related articles