Redaction the AI suggests, and a reviewer signs off.
AI Redaction reads each document and proposes masks for FOIA exemptions, CUI, PII, and any category your legal team configures, each with the exemption code and the reasoning behind it. Nothing is applied until a reviewer accepts, edits, or rejects it.
Every redaction is the reviewer's call.
AI Redaction is an AI layer on top of Feith's Find-and-Redact, which has been in FOIA production for years. It reads each document and proposes masks, but it only proposes: a human reviewer signs off before anything is applied.
- Suggestions only, never automatic application. The reviewer is the deciding party on every redaction.
- Exemption definitions are editable by your legal team. The (b)(3) statutes one agency withholds under aren't the ones another cites.
- Exemptions can be switched off. If your agency doesn't use (b)(1), it never appears in the workflow.
- Works beyond FOIA. The same suggest-then-approve flow covers CUI, PII, and any sensitive-content category you configure. Broad CUI detection across everything at ingest is a separate capability.
On March 14, the investigator interviewed Jonathan Reeves, DOB 1972-08-04(b)(6), (b)(7)(C) at the worksite, corroborated by badge #82-447(b)(7)(C).
Detected, reasoned, and held for review: nothing released on its own.
AI Redaction reads each document and consults your agency's own exemption set to propose masks. Every suggestion waits for a reviewer; their decision is what releases the record.
Why it's different
The reviewer stays in charge
Nothing is masked automatically. A person signs off on every redaction, and that decision is what releases the record.
Every mask shows its reasoning
Every suggested mask arrives with its exemption code, a confidence score, and the reasoning behind it: the legal test applied, the citations to controlling authority, and the analysis of each element. The redaction is ready to defend on appeal.
Tuned to your agency
Your office decides which exemptions apply, so the workflow matches how your office reads the statute.
One review flow, many categories.
FOIA is where most agencies start, but the same suggest-then-approve flow handles every kind of sensitive-content review.
(b)(1)–(b)(9), including (b)(6) personal privacy and (b)(7)(C)
Accept, edit, or reject each suggested mask.
CUI Registry categories and the markings that carry them
The same suggest-then-approve flow.
Names, dates of birth, SSNs, badge and case identifiers
The reviewer confirms before release.
Any sensitive-content rule your legal team configures
Exemption definitions are editable per agency.
Five more capabilities, each its own deep-dive.
These are the other capabilities in the AI in Feith family — separate tools that each share the same permission model, the same audit trail, and your choice of large language model.
Record Insight
Categorize records against your retention schedule, with the reasoning shown.
See the deep-dive 02Prompt Runner
Run custom AI prompts across record sets for extraction, routing, and reasoning.
See the deep-dive 03AI Dashboards
Build dashboards and analytics on records activity from a plain-English brief.
See the deep-dive 04Ask Me
Chat with your records: permission-aware, and cited to the source.
See the deep-dive 06Feith MCP
Connect Claude, Copilot, OpenAI, or in-house agents to your Feith content.
See the deep-diveRelease faster, with the reviewer still in charge.
Bring a sample FOIA case or a sensitive document set. We'll run AI Redaction live and show the reasoning behind every suggested mask, and the reviewer's call on each.
Request received.
Thanks. We'll reach out to set up your AI Redaction walkthrough.