Does SOC 2 require a penetration test?
No. SOC 2 does not formally require a penetration test, but auditors commonly expect one, according to Baker Tilly's guidance on penetration testing for SOC compliance. The criteria ask you to show that your controls work. A penetration test is one of the most direct ways to show it, so most teams bring one.
The reason is simple. SOC 2 is built on the AICPA Trust Services Criteria, which describe outcomes rather than a checklist of tools. Your auditor decides whether your evidence supports each criterion. A test that attacked your product and recorded what happened is evidence an auditor can read, date and trace to a fix.
Your customers ask too. Enterprise security questionnaires often request a recent penetration test report alongside the SOC 2 report. One test, run well, answers both.
Which Trust Services Criteria does a pentest support?
A penetration test is most commonly tied to two of the common criteria: CC4.1, which covers monitoring activities, and CC7.1, which covers detecting vulnerabilities. Neither makes penetration testing mandatory. Both ask for evidence that you evaluate your controls and find weaknesses, and a pentest supplies that evidence directly.
Your auditor maps evidence to criteria, and firms differ in how they do it. Ask yours early which criteria they plan to tie the test to. Then hand over a report that speaks to them plainly.
- CC4.1, monitoring activities: you perform ongoing or separate evaluations to confirm your controls are present and functioning. The AICPA's points of focus list penetration testing as one type of evaluation you can use.
- CC7.1, system operations: you use detection and monitoring procedures to identify configuration changes that introduce new vulnerabilities, and exposure to newly discovered ones. Recurring testing shows that procedure running.
- Remediation: findings that are fixed and re-tested show your process for responding to identified weaknesses at work, with dates attached.
What do auditors accept as penetration test evidence?
Auditors look for a report that shows what was tested, when, what was found and what you did about it. In practice that means a defined scope, test dates that fall inside the audit period, findings ranked by severity, and evidence that serious findings were fixed and re-tested.
Scope mismatches cause the most back and forth. If your system description covers a web app, an API and an AI assistant, a test of the marketing site carries little weight. Test what the report describes.
A Grapefruit report contains the tested hostnames, the run date, every finding ranked critical, high, medium or low, and the captured request and response that proves each one. Your engineer can replay that request before the fix and after it. The before and after is your remediation evidence.
- Scope: the hostnames, applications and APIs tested, matching the system described in your SOC 2 report.
- Dates: when testing ran, inside the period under audit.
- Findings: each issue with a severity rating and enough detail to understand the impact.
- Remediation: what was fixed and when.
- Retest: proof the fix holds, from a later run against the same target.
When should the pentest happen for Type II?
For a Type II report, run the penetration test inside the audit period. Baker Tilly notes that for Type II the testing has to be evidenced within the period under review, so a report dated before your window opens leaves a gap. Test early in the window, so fixes and retests land inside it too.
Timing is where traditional engagements get tight. Stingrai reports that traditional engagements commonly take 2 to 4 weeks from accepted quote to start, 5 to 10 business days of testing, and another 1 to 2 weeks for the report. That is four to eight weeks before you hold a report, with remediation and a retest still ahead.
Start the clock in your favor. With Grapefruit you verify your domain with one DNS record, sign the testing authorization online, and testing starts. Results land within hours of kickoff, which leaves the rest of the window for fixing and re-testing.
Type I vs Type II: what testing evidence do you need?
Type I reports on control design at a single point in time, so one recent penetration test with remediation evidence usually fits. Type II reports on how controls operated across a period of months, so the evidence needs dates inside that period and is strongest when it shows testing as a recurring activity.
Many teams go Type I first, then Type II. A subscription that starts before the Type I date keeps running into the Type II window, so the history is already building when the period opens.
| Evidence | SOC 2 Type I | SOC 2 Type II |
|---|---|---|
| What the auditor evaluates | Control design at a point in time | Control operation across the audit period |
| Test timing | Recent, before the report date | Dated inside the audit period |
| Number of tests | One test usually fits | At least one, and recurring tests show the control operating |
| Remediation evidence | Findings with the fix or plan for each | Fixes and retests dated inside the period |
| What strengthens it | Scope that matches the system description | A dated history of new and fixed findings over time |
How often should you pentest for SOC 2?
At least once per audit period is the common baseline, which for most companies means once a year. Products that ship changes every week benefit from testing more often, because an annual test describes one release, and CC7.1 is about catching the vulnerabilities that new changes introduce.
Consider what changes in a year. New endpoints, new roles, a new AI feature, a new tool handed to an agent. Each one can open a path the last report did not see. A weekly or monthly re-test follows the product instead of the calendar.
Finding issues is half of it. Cobalt's 2025 State of Pentesting data, as reported by CSO Online, shows 48% of pentest findings get fixed, 69% of serious ones, and only 21% of findings in generative AI apps. Recurring tests keep open findings in view until they close.
How does a dated weekly record cover a Type II window?
A weekly re-test produces a dated run for every week of the audit period. Each run lists the findings open at that moment, what is new since the last run and what got fixed. Together the runs form a continuous record of testing and remediation across the whole Type II window.
When your auditor samples a month, you open that month. When they ask whether a critical finding was fixed, you show the run where it appears and the run where it is gone. The answers are already written down, with dates.
- A dated entry for every weekly or monthly run, kept in the portal.
- The verified domains each run tested.
- Findings ranked critical, high, medium or low, each with captured request and response evidence.
- What is new and what was fixed since the previous run.
- A downloadable report for your auditor and your customers.
What should the test cover if your product ships AI?
Cover everything in your system description, including the AI. If your product has a chat assistant, a RAG feature or an agent with tools, those sit inside the same boundary as your web app and API. A test that skips them leaves the newest part of your system without evidence.
Real attacks chain across the layers. A crafted message convinces a support assistant to call an internal API with another customer's ID, and the API returns the record because it trusted the assistant. That is a prompt injection that ends in broken object level authorization. Grapefruit attacks the web app, the APIs, the LLM features and the agents together and reports them in one place.
How do you start SOC 2 penetration testing with Grapefruit?
Add your hostname, publish one DNS TXT record and sign the testing authorization online. It takes a few minutes. Testing starts right after, using test accounts you supply for authenticated coverage, and results land in your portal within hours of kickoff. No demo call, no scheduling.
Pick a full penetration test at a fixed quote for a point in time report, or a maintenance plan for weekly or monthly re-tests through your audit window. Maintenance prices are public and checkout is by card. See pricing, or read the sample report to see what your auditor will read.
Questions
- Is a penetration test mandatory for SOC 2?
- No. SOC 2 does not formally require a penetration test. Baker Tilly notes that auditors commonly expect one anyway, because it is direct evidence that your controls hold up under attack. Most teams include a pentest report in their evidence, and enterprise customers frequently ask for it alongside the SOC 2 report.
- Will a vulnerability scan satisfy my SOC 2 auditor?
- A scan and a penetration test answer different questions. A scan lists known weaknesses. A penetration test exploits them to show real impact. Vulnerability scanning fits CC7.1 well, and auditors commonly expect a penetration test in addition. Ask your auditor what they want to see, then bring both where you can.
- Does the pentest need to fall inside the Type II audit period?
- Yes. For Type II, the testing has to be evidenced inside the audit period, according to Baker Tilly. A report dated before the window opened does not show the control operating during it. Run the test early in the period so the fixes and the retests carry dates inside the window as well.
- How often do I need a pentest for SOC 2?
- Once per audit period is the usual baseline, so once a year for most companies. If you ship weekly, test more often. A weekly or monthly re-test gives you a dated run for each interval, shows what is new and what got fixed, and keeps your evidence current as the product changes.
- What should a SOC 2 penetration test report include?
- Scope, dates, findings and remediation. The report should name the hostnames and applications tested, show when testing ran, rank each finding by severity with evidence, and be backed by a later retest showing serious findings closed. The scope should match the system description in your SOC 2 report.
- Can I hand a Grapefruit report to my auditor?
- Yes. The report is downloadable and lists the tested hostnames, the run date and every finding ranked critical, high, medium or low with the captured request and response behind it. Your auditor decides how it maps to the criteria, so share the sample report with them before your window opens.
- Does a SOC 2 pentest need to include AI features?
- If the AI feature is part of the system your SOC 2 report describes, test it. Chat assistants, RAG features and agents with tools handle customer data and call your APIs. Grapefruit tests them in the same run as your web app and APIs, so one report covers the whole boundary.