← All posts
SOC 2 production data in stagingSOC 2 test dataSOC 2 nonproduction environment

SOC 2 and Production Data in Staging: Decide Before You Copy

SOC 2 does not impose a blanket ban on production data in staging. Map the copy, limit access, record the decision, and keep evidence of the controls you operate.

A staging copy of customer data needs a documented path, owner, access boundary, and end date.
A staging copy of customer data needs a documented path, owner, access boundary, and end date.

SOC 2 does not impose a blanket ban on using production customer data in staging. But a copy creates another place where people and systems can access that data. First ask whether synthetic data or a tested transformation can support the work. If you still need real data, document the reason, who approved it, where it moves, who can reach it, how long it stays, and how you will verify its removal. Review that plan against customer commitments and your proposed SOC 2 scope with an independent CPA firm.

What does SOC 2 ask about production data in staging?

The AICPA Trust Services Criteria include CC6.7, which addresses authorized movement and protection of information during movement. They do not prescribe one test-data method. Management designs controls for its service commitments, system requirements, and risks. The CPA firm examines the agreed system and decides whether the controls and evidence are sufficient.

The AICPA Description Criteria ask management to describe the system it actually operates, including the relevant data and procedures. If staging contains a copy of customer data or can affect production, show that path when discussing the system boundary. Do not treat a label such as “test” as proof that the environment is outside the examination.

This also matters before a report is planned. A customer questionnaire may ask where its data goes. Answer from the current system map, not from a future policy. The SaaS scope guide helps connect the data path to the service customers use.

Choose the least sensitive data that answers the test

Start with the defect or behavior you need to reproduce. What fields and relationships matter? Can you build synthetic records with those properties? If a transformed copy is needed, test whether the transformation preserves the required behavior while removing customer identifiers and sensitive content. Check free-text fields, attachments, logs, search indexes, and linked records; changing a name column alone may leave a person identifiable.

The OWASP API Security guidance on inventory advises against production data in nonproduction API deployments and says unavoidable use needs production-grade security treatment. NIST’s Secure Software Development Framework, PO.5.1 calls for development environments to be separated and protected. These are engineering guides, not extra SOC 2 criteria. Use them to challenge the design, then document the control your team actually chose.

Test need First option to try Decision to record
A known edge case Synthetic records with the needed shape Which fields and relationships reproduce it?
Realistic volume Generated data or a transformed sample Does the transformation leave identifiers or sensitive content?
A customer-specific defect A narrow, approved data extract Why is real data needed, who may access it, and when will it be removed?

Do not claim a transformed set is anonymous until you have checked what remains and what other data a recipient can join to it. A test environment with real customer data needs controls that fit the actual exposure and your commitments, even if it never serves a public request.

Map one copy from source to deletion

Write down the path before running the job. The data flow diagram guide gives you a way to show every system and recipient. For a proposed copy, record these facts:

  1. Record the test case, requester, approving owner, and decision date. Check customer terms before moving data.
  2. Name the production dataset, export tool, transfer path, staging account, storage, backups, and any external processor.
  3. List the data classes, fields removed or changed, transformation version, and a check of the resulting sample.
  4. Identify people, service accounts, contractors, and agents that can read or export the copy, plus review and logging sources.
  5. Set an expiry or refresh rule, deletion owner, deletion record, and treatment of backups or derived copies.

An old staging database may already contain data that nobody remembers copying. Inventory it before declaring that only synthetic data is present. Compare the claimed access boundary with actual grants and recent use. The access review evidence guide shows how to record who reviewed access and what changed. If the copy reveals a gap, record the correction and date rather than rewriting the past.

Keep evidence that the decision operated

A policy saying “production data is not used in staging” cannot support that claim when a refresh job copies the database every week. Keep source records for the actual path: approval, job run, destination settings, access grants, transformation test, expiry check, deletion result, and exceptions. For a Type 2 examination, dated records need to cover the relevant period, not just the configuration visible on the day someone asks.

Keep the GRC record lean. It can hold an opaque dataset or request ID, owner, decision, dates, scope, and links to controlled evidence. Avoid committing real customer records, secrets, raw exports, or personal data that may need erasure into Git. Store sensitive source evidence under access and retention rules that fit it. The evidence repository guide and retention guide explain that boundary.

FileGRC is a Git-native GRC workspace for SOC 2 work. In the FileGRC workspace, JSON holds structured records, Markdown holds long-form work, and Git supplies the change history. Your team can link the data-flow decision, risk owner, review, and evidence references so an agent can find stale copies or missing approvals in the records it can access. Starter records are proposals, not compliance claims. The database, identity, cloud, and testing systems perform the copy, protection, and deletion. Your team reviews the facts and customer answers; an independent CPA firm performs the examination and judges evidence sufficiency.

Open source · MIT

Run your SOC 2 program as files in Git.

Keep policies, controls, work, and evidence indexes in a repository your team and agents can inspect. Add optional FileGRC Autopilot email and Slack reminders to keep work moving.

Frequently asked questions

Does SOC 2 ban production customer data in staging?

No blanket SOC 2 rule bans it. Management must understand the data path, commitments, risks, and controls it actually operates. Prefer synthetic or suitably transformed data when it can answer the test. If real customer data is needed, authorize and protect the copy, record its location and lifetime, and discuss the proposed control approach with the CPA firm.

Is masked production data safe to use in a test environment?

Masking can reduce exposure when the transformation fits the data and the remaining fields, recipients, and any reidentification keys are controlled. Test the output and its recipients before calling it de-identified. Treat a copy that still exposes customer information as sensitive data and apply the controls your commitments and risk assessment require.

Does staging have to be in our SOC 2 scope?

That depends on how staging affects the service and whether it stores or processes customer data, supports production changes, or can access production systems. Describe the actual path and proposed boundary with the independent CPA firm. Calling an environment nonproduction does not by itself remove its risks or responsibilities.

What evidence should we keep for a production-to-staging copy?

Keep the purpose, approval, source and destination, data classes, transformation result, access list, transfer record, retention limit, deletion or refresh record, and any exception or incident. The source systems should prove the actions; a policy or configuration file alone does not prove that a specific copy followed them.

Can FileGRC mask or delete customer data in staging?

No. FileGRC is a Git-native GRC workspace for SOC 2 work. It keeps structured records in JSON, long-form work in Markdown, and change history in Git. Your data systems perform masking, access control, and deletion. FileGRC can organize decisions and evidence references, while your team reviews the facts and the CPA firm evaluates sufficiency.