AI Smart Redact

Powerful Redaction: AI-Detected, Human-Approved

Pdftools combines non-generative AI and a proprietary rules engine with mandatory human-in-the-loop review to deliver secure redaction of sensitive data.

True redaction with no trace left behind
Self-hosted, scaleable, and with full observability
Up to 98% accuracy, zero risk of hallucinations

Try demo now Contact us

0.054 €/page at standard volumes; better rates available at scale

How AI Smart Redact Works

All documents in the pipeline go through three stages to ensure all final decisions are made by a human before redaction is applied.

Identify with speed

A non-generative AI model and our deterministic rules engine work together to flag potentially sensitive elements.

Decide with expertise

An authorized reviewer on your team evaluates the suggestions and decides which information needs to actually be removed.

Remove with certainty

Other tools use black boxes or partial removal. We apply true redaction, generating a new, clean PDF with no hidden layers or recoverable content.

Robust detection engine with human-in-the-loop

Our proprietary detection engine combines stochastic and deterministic detection to give recommendations for redaction. This makes the process faster and more efficient. But the final decision always rests with the user.

Stochastic detection

The basis of detection is a compact NER model. We don’t use generative AI and achieve better performance, as well as more robust and reliable detection. A confidence score for results creates clarity.

Deterministic detection

Our proprietary rules-based engine complements AI detection to improve accuracy and control. It uses regex and structured data to refine results and deliver transparent, reproducible outcomes.

Human-in-the-loop review

AI Smart Redact displays redaction suggestions to the user for review. Users can also manually add redactions. The final decision is always made by a human for clear accountability and compliance.

Applying redaction

Rather than overlaying or deleting content, AI Smart Redact completely rebuilds the PDF, copying over only what is visible after marking for redaction. Unnecessary data is discarded.

Full privacy for document redaction

AI Smart Redact stands out by offering privacy and security features that highly regulated industries need. We’re uniquely positioned to keep highly sensitive data secure.

Self-hosted vs SaaS

Most competitors offer redaction as software-as-a-service. AI Smart Redact is self-hosted and can be used fully offline.

True redaction vs residual data

Competitors visually overlay sensitive information so the original content can be recovered from the document structure. AI Smart Redact removes sensitive information entirely—no residual data.

Human-in-the-loop vs fully AI-automated

Fully automated redaction comes with uncertainty. AI can detect sensitive content, but only humans can decide what must be removed.

Privacy-first vs traces of PII

Input and output is encrypted and any documents surfaced in the application are deleted after two days by default or as configured.

For platform-agnostic redaction applications

AI Smart Redact uses Docker images and works with all OS. Input and output is handled through APIs or manual upload.

See documentation

Pdftools understands the needs of regulated industries

With 25 years of experience serving customers in industries that handle sensitive data, we understand how important it is to redact PII from different types of documents.

AI pipelines

Neutralising documents to prepare them for LLM training, ensuring privacy and compliance

Banking

Redacting PII from KYC documents to meet GDPR and financial services compliance requirements.

Insurance

Removing PII from claims documents so it can’t be recovered during regulatory reviews.

Ready to take the next step?

See how AI Smart Redact handles sensitive documents in your environment

Book a demo