Blog
Benchmarks, methodology, engineering notes.
The alphabet tax: shrinking our AI to fit a browser tab
Our detection model carried a vocabulary for 104 languages. We ship seven. How we cut 39 MB without touching a weight — and why a browser tab is the hard part.
Read the postFuture of WorkWe built the perfect conditions for hiding AI use
Admit you use AI and your trade calls it cheating; be overheard and your employer bans it. Hiding is rational — and it makes AI policy unenforceable.
Read the postLegalRedact PII from legal PDFs without uploading to the cloud
Online redaction tools put privileged client data on someone else's servers. How to redact legal PDFs on your own machine, with the text truly removed.
Read the postHow it worksReversible redaction: how tokenization un-redacts a file
Reversible redaction swaps each name for a stable code and keeps the map on your machine — share the document, decode it later. How tokenization works.
Read the postComplianceGDPR anonymization: what actually counts as anonymized
Under the GDPR, data is anonymized only if no one can realistically re-identify anyone. Keep a mapping and it is pseudonymized — still fully in scope.
Read the postGuidesHow to redact scanned PDFs and images
A scanned PDF is a picture, not text, so text-based tools find nothing and miss everything. The OCR pipeline that actually redacts scans, on your device.
Read the postAI SafetyIs it safe to upload my contract to ChatGPT?
Most NDAs and commercial contracts restrict who you may disclose them to — and uploading a copy to a third-party service is a disclosure. Read the clause.
Read the postEngineeringWhy we built our own demo studio
We started with an off-the-shelf demo-video tool. Two themes, seven languages and a weekly-changing UI broke it — so we built our own. Why that matters.
Read the postPositioningWhy nobody built a redaction tool for the person holding the scan
Almost every PII tool takes text, not a page — so it starts after the hardest part is done. That assumption is why redacting a PDF full of images is still so hard for everyone who isn't a developer.
Read the postWorkflowThe same name, the same code, every time
Professionals iterate with the AI across many turns. That only works if one entity keeps one code the whole way — conversation-stable anonymization.
Read the postPositioningHow companies protect PII — and the gap we fill
Companies protect personal data four ways: governance platforms, cloud APIs, managed redaction, manual blackout. An honest map, and the gap we fill.
Read the postEngineeringWhy we built our own PII engine instead of shipping Presidio
Presidio is an excellent toolkit. We still wrote our own engine, because we ship a bounded PDF on a laptop, not a text stream in a cloud pipeline.
Read the postAI SafetyIs it safe to paste documents into ChatGPT?
Short answer: sometimes — and which case you're in is entirely under your control. What determines the risk, and the one step that makes it moot.
Read the postResponsible AIA machine can't be responsible. You can
Responsibility means having something to lose when things go wrong, and a machine has no stake to forfeit. Why a human stays in the loop, always.
Read the postData RiskTo those who quietly share client documents with a chatbot
AI makes exposing confidential data effortless, and the consequences are neither theoretical nor distant. What you risk pasting client files into a chatbot.
Read the postPrivacy EngineeringPII protection is different when AI is in the loop
For 30 years, protecting personal data meant controlling database access. AI created a frictionless path from controlled documents to the public cloud.
Read the postBenchmarkspromptShield vs Microsoft Presidio: a PII benchmark
A reproducible head-to-head on 14 PDFs in 7 languages. Presidio's default install emits 666 spans, promptShield 252. Most of the gap is noise, not skill.
Read the post