Job personal-data-finder-script-for-a-file-folder · revised 12 October 2026
Show me which files contain personal data before I use AI on them
A script you run on your own computer lists files that look like they hold personal data, by path and type, never values. Recall is measured on a set we plant and a folder you plant. No guarantee.
You might be seeing
- Nobody can say whether a shared folder holds personal details
- An AI tool is about to be pointed at files nobody has read
No passwords, keys, card details or admin invites needed to start.
What usually happened
Before documents go to an AI tool you need to know which ones hold personal data. Even a widely used open tool says there is no guarantee it finds all sensitive information, so a finder gives a list to look at, not an all-clear. The danger is a buyer reading 'none found' as 'none there'.
Who it’s for: Owner or office manager about to point an AI tool at a shared folder and unsure what is in it.
Usually starts when: A folder of documents is about to be fed to a tool, and nobody knows whether it holds customer or staff details.
The result: A script you run locally lists the files that contain likely personal data, by path and type, never the values. Recall is measured on a planted synthetic set we make (a smoke test) and on a test folder you make, and reported by type. The script makes no network call, which is tested, and its report says plainly that it does not guarantee to find everything.
Check whether this job fits
Answer these without sending files or logins. Nothing is submitted unless you choose to contact us.
What you get
- The script and run steps
- The report format: paths and types, never values
- Recall by type on the held-back planted set
- The no-network test
- A page on what the finder cannot find
- Steps for your own planted-folder test
Included
- A script for one operating system that you run on your own computer
- Folders of up to 50,000 files: text, Word, Excel, CSV and PDFs with a text layer
- Pattern types: email, phone, card-like and ID-like numbers, names and addresses
- A planted synthetic corpus of 200 files, 100 held back with at least 100 planted items of each pattern type, to measure recall
- A report of paths and types only
Not included
- A guarantee that all personal data is found
- Redaction or deletion of anything
- Advice on whether you may process the data
- Image-only files and handwriting
- Sending any file or value anywhere
How we know it’s done
Agreed with you before work starts. Each check produces evidence you keep.
On the 100 held-back planted files, at least 90 of 100 planted items of each of email, phone and number patterns are found (90 of 100 shows only that the true rate is at least 83% with 95% confidence); names and addresses are reported but not passed or failed. We plant these files ourselves, so this is a smoke test of the script. In a test folder you make, with 20 made-up email addresses and 20 made-up phone numbers that you place one to a file inside 20 files of the supported types for each type and do not tell us about, the report lists the file for at least 18 of 20 of each type (at least 71% with 95% confidence). The results describe the planted sets, not your folder.
Evidence: The recall table with the list of misses, and your test folder beside the report.
Run with the network blocked, the script gives the same report; a connection monitor shows no outbound connection.
Evidence: Both reports and the monitor log.
The report holds paths and file types, never values: a search of the report for the planted values finds none.
Evidence: The search output.
The report's first line says it does not guarantee to find all personal data, and unreadable files are listed.
Evidence: The report's first lines.
Running the script does not change your files: checksums before and after are identical.
Evidence: The checksum comparison.
Sign-off. You inspect the recall table, the no-network test and a sample report, and sign off in writing. Payment follows sign-off.
If it fails. If the held-back test is not passed, you do not pay for this fixed scope. You keep the results, the failures and the test set, and we agree whether to stop or re-quote. No surprise work.
When it fits, and when we stop
It fits when
- You can run a script on your own computer, or a colleague can
- The folder has up to 50,000 files of the supported types
- A named person reads the report and decides what to do
- You accept that 'none found' is not 'none there'
- A named person runs the script on a test folder you make, then on the real folder, and reads the report
We stop and tell you if
- You want a guarantee or a compliance certificate
- You want redaction done by the script
- Most files are images or handwriting
- You want your files sent to us to be scanned
What could go wrong
We receive no files from you. The script only reads the folder you point it at and writes a report; it changes no file.
Scroll the table sideways to read it all.
| Risk | How we handle it |
|---|---|
| 'None found' is read as 'none there'. | The first line of every report says it does not guarantee to find all personal data, and image-only files are listed as unread. |
| The test set looks like the script and flatters it. | Half the planted files are held back. Recall on the held-back half is what is reported. |
| The script sends a value somewhere. | A no-network test is part of the acceptance. The report holds paths and types, never values. |
A separate automated reviewer checks the report format, the recall table and the no-network test before hand-over and does not mark the held-back test. A named person on your side runs the script and signs off. No person on our side is included unless your agreement names one, and at launch the work is largely automated.
How we deliver
We arrange the work and independent review, then show you the result against the agreed checks. You keep authority over your systems.
- Agree the file types, the pattern types and the severe errors
- Build a synthetic corpus of 200 files with known planted items
- Build the finder on 100 files and keep 100 held back
- Measure recall by type on the held-back files; run the no-network test
- Independent review of the report format and the recall table
- Hand over the script and run steps
This is a one-off job, not emergency cover or a subscription. We confirm eligibility, the price, a start window and a delivery date before you accept, and work starts only after inputs, a secure handoff route and permissions are in place. Model usage up to a cap agreed in writing is included, and an enquiry creates no charge or booking.
Need to keep it working?
Need personal details masked as well? See the spreadsheet masking job listed under related jobs.
Ongoing work is separately scoped and quoted: no monitoring, response-time guarantee or automatic subscription is included in this job.
Explore an ongoing managed lane, or mention the responsibility you need in your enquiry.
What you can check
This is a new service. We have not delivered this job for a client yet.
Other ways to get this done
- Microsoft Presidio is a free open-source toolkit for finding and masking personal data. It needs someone who writes code, and it also gives no guarantee. presidio.dataprivacystack.org
- For a small folder, searching for '@' and digit patterns with your computer's search tool finds the obvious cases.
Questions
Can I trust a clean report?
No. A clean report means the script found nothing it knows how to find. It is a list to look at, not an all-clear.
Do my files come to you?
No. We build the script on our own synthetic files. You run it.
Does the recall figure apply to my folder?
It applies to our planted test set. Your folder may differ, which is why the report is never an all-clear.
Send an enquiry
Send us
- The kinds of files in the folder and roughly how many
- Which operating system you would run the script on
- No real records, personal data or credentials in the first enquiry
Later, once you agree
- Nothing from your folder: we build and test on our own synthetic files
- A named person who runs the script and reads the report
Your files never leave your computer. We build and test the script on synthetic files we create, hand it over, and you run it yourself. We ask for no access to your folder.
Email fallback: open your mail app
If website submission is unavailable, review and send the fallback email yourself. An email fallback is not a website receipt. Or write to hello@syntheticindustry.ai with “personal-data-finder-script-for-a-file-folder” as the subject.