redactX: Information Protection, Transparency, and Information Sharing: a growing natural tension

Government organizations at many levels (federal, state, local) generate, collect, share, and work with sensitive information that is controlled and needs to be protected from unauthorized disclosure. There are also rules, regulations, and processes that govern the release of such information in a “sanitized” form to the public. Examples include Freedom of Information Act (FOIA) requests, Health Insurance Portability and Accountability Act (HIPAA), legal court requests, Mandatory Declassification Reviews (MDR), etc. With growing volumes of data and stringent privacy regulations, secure and accurate redaction has become essential for organizations in law enforcement, government, healthcare, and legal services. Traditional manual tools often slow the redaction process, increasing the risk of errors and inefficiencies.

redactX, founded by Joseph Santucci, is an AI-powered platform designed to automate redaction across a broad range of modalities: documents, videos, images, and audio files. Automation brings benefits at many levels: it reduces the amount of work and time needed to produce a redacted output file, it reduces the possibility of human errors, it supports consistency in what is being redacted across different workflows and activities, etc. Automation, however, can only be truly achieved if the software provides very high detection accuracies, and this is where the redactX software makes a huge leap for the industry.

Breaking Obstacles with a Smarter Alternative

Conventional redaction tools envision users interacting with a document or a video on a frame-by-frame or sentence-by-sentence basis marking manually what is to be redacted. The process is then repeated on every new file. Sometimes there are limitations on what type of source material can be processed (ex: only body-worn video from a specific camera type) or where the processing takes place (ex: the file must be uploaded to a vendor operated cloud), or how much material can be processed (ex: per-minute of audio file charges), or the scope of what the redaction tool covers (ex: different applications needed for pdf, video, audio, etc.), and so on.

redactX flips the equation around. Users define redaction criteria, and the software implements these criteria to automatically generate a redacted output file. Criteria can be saved and reused to process batches of files or as part of automated tasks that occur on schedules for different types of documents. For example, a criteria might be “redact all faces in this folder of videos” or it might be “redact only these two faces and leave the rest visible in all the images, videos, or pdf documents at this location”. With a high degree of accuracy, the redactX software detects object, faces, text strings, entities, speech, etc. in many different file formats, and according to the criteria specified by the user. A few clicks of the mouse let users accomplish what used to take hours and days of work. Several different types of advanced AI technologies are used to accomplish this, including Computer Vision, Natural Language Processing, Speech Recognition, Machine Learning, etc. The application’s UI/workflow focuses user attention on areas that require that attention (drill-down capabilities, different views of the redaction data, reports, “one-click” error correction, etc.). The level of functionality provided is significant. For example, the software can detect/recognize, and therefore redact, concepts (we call these “entities”). Things like “dates”, “names”, “locations”, “events”, etc. are entities. In addition to an extensive built-in list of entities, the software lets users create their own definitions (“zeroshot” entity recognition) without needing to retrain the system. This level of adaptability is unique in the industry.

The software can be installed on either customer premises or in private clouds. There are no external connection requirements, and everything is controlled by the customer. The data is under full control of the customers. There are no restrictions on the source of the files, or quantities. The software is suitable for deployment in air-gapped protected networks, a typical government customer requirement.

“Our objective is to save time and labor through a platform that automatically redacts with minimal user input” says Thomas Scanland, SVP at redactX.

Redaction Workflows: the impact of automation

redactX has transformed redaction work across multiple industries. With customers in Federal Government, Law Enforcement, and Healthcare, the software has proven itself in many different application spaces and workflows. In all cases the objective has and will continue to be “do more and better with fewer resources”. For instance, in one DOD application the goal was to help legal teams respond to court ordered requests of pdf documents that contained both CUI (Controlled Unclassified Information) and PII (Personally Identifiable Information). Dealing with an increasing volume of documents, the challenges of internal requirements and FOIA requirements, and accelerated court schedules necessitated the functionality of the redactX software. In another police application the automatic redaction of faces of young children in evidentiary videos was a big time saver. In yet another application, working with the American Board of Surgery (ABS), over 300 surgeons submitted endoscopic procedure videos for recertification purposes, where the videos were automatically scrubbed by the redactX software to remove any PII information (faces of patients and nurses, text on whiteboards, etc.). The feedback from the doctors was excellent, and overall the use of video for ABS certification reviews would not have been feasible without the redactX software.

Redaction Workflows: the future of automation

It is clear that the need for automating redaction workflows will only increase going forward. The natural tension created by Information Protection, Transparency, and Information Sharing requirement is here to stay. Advanced AI technologies such as the ones provided by redactX are key in this environment. redactX is a beacon in the industry as it continues to develop its software to provide progressively higher levels of functionality and automation for its customers. We can expect more embedded human-level understanding of content by the software which will translate directly to less work for the user.

For more information, please contact us at: tscanland@redactx.com