Glossary. Shadow data.
What is shadow data, and how is it different from shadow IT?
Shadow data is sensitive data in copies nobody tracks: restores, exports and extracts. It can sit in an approved system, so the fix starts with an owner.
By Dipal Panchal, co-founder and CEO · Updated November 8, 2026
Copies outside the list
Shadow data is sensitive data held in copies, exports or systems outside the ones your teams track, such as a test restore or a spreadsheet export. Unlike shadow IT, which is an unapproved tool or service, shadow data can sit inside approved systems, in places nobody listed. A copy with no owner is shadow data wherever it sits.
Someone needed realistic data for a test, a quick extract for a report or a file to send a colleague, and the copy outlived the reason.
Six copies, and how each was found
Copies on one estate in four markets, from a generic scan, not IRIS output.
| Source | Location | Detected type | Why | Sensitivity | Decision |
|---|---|---|---|---|---|
| Data warehouse, UK | analytics.customers_full_2024 | Email, phone number, date of birth | Found by a warehouse scan; the loading team says it is a full CRM copy from a 2024 campaign, with no owner | Confidential | Owner named; the copy is deleted once the campaign record is kept |
| Test database server, India | qa_restore_0412.customer_kyc | PAN (Indian tax ID), Aadhaar number | Found by a scan; the database team says it is an April production restore kept after load testing | Restricted | Replaced with synthetic values; dropped once no hold or retention duty applies |
| Microsoft 365, US | Finance > Exports > card_refunds_q3.xlsx | Names, payment card numbers | Found by a OneDrive scan; finance says it was emailed round, then saved | Confidential | Moved back into the finance system; the file is deleted |
| BI extract bucket, US | bi-extracts/daily_patients.csv | Patient names, appointment dates | Found by a scan of the bucket; the BI team says it refreshes nightly from the EHR | Restricted | Kept with a named owner; access narrowed to the BI team |
| Support desk (SaaS), UAE | Ticket attachments | Passport scans | Seen in a manual review; the support desk isn’t connected to any scan | Restricted | Excluded from the scanned count; listed as not scanned until connected |
| A developer’s laptop | Downloads folder | Unknown | Not scanned: no scan in this estate reaches local folders | Unknown | Excluded and listed as a gap; synced cloud folders are scanned instead |
How copies end up there
Rows 1 and 4 follow a path Dipal Panchal, Sylox’s co-founder and CEO, describes from enterprise estates, where data leaves the system that collected it for a job. On a call in October 2026 he said, “We create read replicas of that database. We put it in our warehouse, lakehouse, analytics, in CRM, cross-sell.”
Row 2 is the kind he names from CISO conversations. On a call in 2026 he said test and development databases fall outside the reports a CISO receives. GDPR Article 32 still applies to personal data in a test database.
Shadow IT is a tool; shadow data is a copy
Shadow IT is a tool or service the organization didn’t approve, such as a file-sharing app bought on a company card. Shadow data is the sensitive data that ends up somewhere untracked, approved system or not.
In this estate, rows 1 to 5 sit in approved systems: a warehouse, a test server, OneDrive, a BI bucket and the support desk.
Every copy gets an owner and an end date
Our rule: every copy of sensitive data outside its system of record gets a named owner and an end date when it is made. A copy found without either is assigned to the owner of the source it came from, who decides within one review cycle to keep it, replace it with synthetic data or delete it. Deleting includes the residue: recycle bins, versioned buckets, time-travel tables and backups.
Test restores get the shortest end dates. A load test needs realistic shapes and volumes, and values provide both without the real records.
What a scan for shadow data misses
A scan finds copies only in the sources it reads. Rows 5 and 6 show the gaps: a SaaS tool nobody has connected yet, and local folders on laptops, which no scan in this estate reaches.
A scan also can’t say why a copy exists or whether it should. It shows a full customer table in the warehouse; the team that loaded it has to say whether the campaign it served is over.
IRIS finds sensitive data wherever it sits among the sources you connect, each finding with its scan date; your team decides which findings are copies. Capabilities vary by edition.
Sensitive data in connected databases, warehouses, cloud storage and Microsoft 365: IRIS produces part of this, in the sources you connect, for the data types IRIS detects.
Who can reach each copy: IRIS produces part of this, where the source exposes permission data.
Copies in SaaS apps and other systems nobody connected: Not produced by IRIS.
Deleting or moving a copy: Not produced by IRIS.
Explore discovery and classificationList recent production restores
Ask your database team for every non-production database restored from production in the last 12 months, from change tickets or the restore log where the platform keeps one. For each one, write down the owner and the end date; any restore with neither is your first shadow data finding, and it needs a decision before the next release cycle.
Copies found this way belong in your sensitive data inventory. For the full method on test environments, see how to find production personal data in your test databases.
Finding copies, and keeping them listed
What is sensitive data discovery?
The scan that found rows 1 to 4, and the record of what it couldn’t read.
What is scan coverage?
The record of what a scan read and couldn’t; sources nobody connected need a list of their own.
What is a sensitive data inventory?
Where each copy in the table goes once found: a dated list with an owner and a note of how it is known.
What data teams ask about copies
No. Dark data is collected and often never used. Shadow data is untracked, and it may be used every day, like the nightly BI extract in row 4.
No, not when they are inventoried with an owner and a retention period. A test database restored from one of those backups is a new copy, and it needs its own owner.
Under our rule, the owner of the source it was copied from, until the copy is deleted or someone else accepts it in writing.
Protect your brand. Start with your data.
Pick the warehouse or test server you trust least. We’ll show you where IRIS finds sensitive data in it.