Omit Systems
We build Omit, software that finds and redacts personal data on your own machine. Nothing is uploaded, there is no account to create, and there is no telemetry being collected.
Omit Systems exists because of a gap we hit building something else. We worked on Telos, an app where people write down the most personal things about themselves, and making it useful meant sending some of that writing to a language model. The exposure was never a hacker and never a stolen laptop. It was ordinary product work, moving private text to a service we did not control, with nothing off the shelf that would strip the personal details out first without uploading them to do it.
You cannot train, warn or govern your way out of that at scale. Every policy is a request that a busy person has to remember to honour, every single time, and that is a hope rather than a control. The fix that holds is structural: make the sensitive data never leave the device. If nothing goes out, there is nothing to leak, no matter what anyone pastes or forgets.
What we hold ourselves to
Offline is the architecture, not a setting
Detection and redaction run on the machine in front of you. There is no server on the other end of a network call, because there is no network call. You do not have to take that on faith: disconnect the machine and redact a file end to end. If any part of Omit needed to phone home, that is where it would fail.
Your original file is never modified
Redaction writes a new file and leaves yours exactly as it was, in every format. The write itself is a temp file followed by an atomic rename, so a crash midway cannot leave a half-written file where your data used to be.
Uncertain means refuse, not release
Detection is statistical, and recall on arbitrary prose is never one hundred percent. Any vendor claiming otherwise is overselling. Omit fails closed, redacting when it is unsure rather than letting a value through, and a review step is available on every run.
Claims get measured, including the ones we lose
Where a claim can be measured it is measured, and the corpus and the metric are published alongside it. The comparison against Google Cloud DLP started with Google ahead on nine of fourteen entity types in ordinary prose. That result is published too, along with the three defects in our own detection that chasing it uncovered.
Who builds it
Omit Systems was founded by Ranjan Singh, who writes the founder posts and the benchmarks on the blog. Product reference and documentation ship under the Omit name rather than a byline.
The security page sets out the offline guarantee and how to verify it for yourself. Omit Redact is the application that ships today, and Protect, the clipboard guard, is free for everyone permanently. For anything else, email hello@omitsys.com.