Redaction That Runs Without Anyone Opening the File

Classes are selected, the asset is processed, and a redacted rendition is produced, with no operator opening the file. Selectable per job: the object and PII classes to detect, the masking style, the confidence threshold, and the output policy that decides what happens to the original.

Review remains available. Automation is the default path for volume, not a replacement for judgement on a consequential release.

Bulk, permissioned per format

Redaction runs across a set of files rather than one at a time. The permission to do so is granted per format, so an operator can be authorized for bulk document redaction without being authorized for bulk video. Audio, video, document and image are four separate rights.

Processing is queue-based and runs unattended, so a batch submitted at the end of a day is worked overnight.

Volume has been exercised at over 1.1 million recordings.

Templates for recurring document types

Where documents share a layout, the regions carrying personal data sit in the same places. A template records those regions and the classes to detect, and applies them to every document of that type.

Bulk redaction takes a template, which is what makes a queued batch consistent rather than dependent on which operator submitted it. A form that arrives every week is processed the same way each week.

API

A REST API with an OpenAPI specification and OAuth client-credentials authentication, for upload, processing, redaction and retrieval. Two shapes are common:

• Fully automated — upload, trigger analysis, start the job, receive status by webhook, retrieve the output. Nothing human in the loop.
• Hybrid — upload by API, AI detects, analysts review in Studio Space, your system retrieves the approved output.

• FOIA and public records — surge volumes after a high-profile incident
• Call recording redaction — thousands of recordings per batch
• Call centers — continuous inbound volume
• The API — embedding redaction in your own product

• Automatic redaction applies what the detector found at the configured threshold. Where a release is consequential, detection review is the step that catches what the threshold admitted or missed.
• Bulk processing is queued work; throughput follows available AI processing capacity.
• A template assumes a stable layout. A document whose structure differs still needs review.

• An automated pipeline runs redaction over selected classes without an operator opening the file.
• Hundreds of files redact at once as queued bulk work, running unattended overnight.
• Bulk redaction is permissioned per format: audio, video, document and image are separate rights.
• Reusable templates capture the redaction pattern for a recurring document type and apply on their own or across a bulk batch.
• Templates are what make batch output consistent between operators.
• Volume has been exercised at over 1.1 million recordings.
• REST API with an OpenAPI specification and OAuth client-credentials authentication.