Does SOC 2 require SSO? What is tested instead, and the path without an identity provider
Updated
No. None of the five sources fetched for this page says the SOC 2 criteria name single sign-on. SSOJet, a company that sells SSO, writes that "SOC 2 does not name single sign-on as a mandatory control". What an auditor tests is the result: who can sign in, how access is given and taken away, and whether you followed your own written rules. SSO makes those easier to show. A small team can also show them with a list of accounts, MFA in each application and an offboarding checklist.
What the sources say about SSO
Most pages that answer this question are written by companies that sell SSO. Five pages were fetched for this guide: two vendors, two articles from one audit firm, and one accounting firm's advisory team. None calls SSO a requirement.
| Source | Who they are | What it says about SSO |
|---|---|---|
| SSOJet (Avi Kapoor, 24 July 2026) | Sells SSO, MFA and SCIM for SaaS products | "SOC 2 does not name single sign-on as a mandatory control. It requires the outcomes in CC6.1 through CC6.3". Calls SSO "the most practical way to meet those outcomes" |
| Secureframe (no byline or date shown) | Sells a compliance platform | Lists "Implementing single sign-on (SSO)" among practices organizations can adopt "to exceed SOC 2 requirements", and notes that "the points of focus are not requirements" |
| Linford & Co (Isaac Clarke, CPA, partner, 17 April 2024) | Audit firm that performs SOC 2 examinations | Lists SSO's benefits, among them "Centralized control can be easier to manage than many separate logins." Does not say it is required |
| Linford & Co (Hilary Stavrakas, CISA, partner, 21 May 2025) | Same firm | Lists "Single Sign-On (SSO) to streamline and centralize access management" as one way the access process "could be enhanced", next to MFA |
| Boulay (Risk Advisory Team, 17 January 2024) | Accounting firm; the team offers SOC 2 services | "Implementing SSO aligns closely with several SOC 2 criteria, particularly those related to security." Aligns with, not required by |
Sources: (SSOJet) (Secureframe) (Linford, identity and access management) (Linford, access control maintenance) (Boulay). Two limits. The Linford articles are about access management in general, not about this question, so their silence on a requirement is weaker evidence than a direct statement would be. And nobody here quotes the AICPA's criteria text on SSO, because by these accounts there is none to quote. Your CPA firm has the last word for your system.
What is tested instead
SSOJet maps the sign-in part of an examination to four criteria. This is one vendor's mapping; the criteria names in the first column link to this site's pages on each.
| Criterion | What the auditor checks, per SSOJet | What you show without SSO |
|---|---|---|
| CC6.1 | "Access is restricted and authenticated before it is granted" | The password and MFA settings of each application, one screenshot or export per application |
| CC6.2 | "Users are verified on registration and removed when access ends" | An onboarding ticket and an offboarding checklist per person, with dates |
| CC6.3 | "Access follows roles, least privilege, and segregation of duties" | A role matrix and the signed periodic access review |
| CC7.2 | "Access events are logged and anomalies are detected" | Sign-in logs from the cloud, source control and the main business applications, and who reviews them |
The middle column is quoted from SSOJet's table (SSOJet). The right column is this page's reading of what a team without an identity provider would hand over, not a list from any source. Secureframe makes the general point behind it: the criteria come with points of focus, and "the points of focus are not requirements" (Secureframe). The same logic applies to MFA, which is covered in does SOC 2 require MFA?
What SSO makes easier to show
- One place to switch a leaver off. Boulay: "When an employee joins or leaves company, administrators can update access permissions in one central location, ensuring that changes are reflected across all connected systems."
- One MFA setting. Enforce it at the identity provider and every application behind it inherits it. Without SSO, each application has its own setting and its own export.
- One sign-in log. Boulay says SSO systems "provide robust audit trails" of user activity (Boulay).
SSOJet pushes the first point hardest. It calls CC6.2 the place "where most teams fail" and writes "Manual offboarding almost never survives this test", with the example of a ticket filed days after someone left (SSOJet). It sells the automation that fixes this, and it gives no data for "most" or "almost never", so treat those as a seller's argument. The mechanism it describes is real, though: in a Type II examination the auditor looks at leavers across the whole period, and one late removal is one exception. How leavers are sampled is in SOC 2 offboarding evidence.
SSO also has a limit that the sellers mention less. Applications with their own user tables, tools joined with a personal email and API tokens sit outside it. Even with an identity provider you still need the list of what is not behind it.
The path without an identity provider
The generator asks which identity provider you use: Google Workspace, Okta, Microsoft Entra, JumpCloud or none. Policyseed sells a kit, so read the text below as one example of how to write this down, not as a requirement. The excerpts are not typed by hand. They are rendered from the policy templates for the Northwind Cloud Inc sample intake twice, once with its identity provider (Okta) and once with "none", so they change when a template does.
Where identity lives (Access Control Policy, 4.3)
With an identity provider: Okta is the single source of workforce identity. Business applications are integrated with it for single sign-on and, where supported, automated provisioning and deprovisioning; applications that cannot federate are recorded in the exception register with a compensating review.
With none: Until a central identity provider is adopted, the Security Owner maintains a master list of every application account for every person, and adoption of a central identity provider is a tracked security objective with a committed date.
How people sign in (Authentication and Password Policy, 4.3)
With an identity provider: Applications that support single sign-on are integrated with Okta, and personnel sign in through it rather than with application-local passwords, which are disabled where the application allows.
With none: Until a central identity provider is adopted, each application account has its own unique password and multi-factor authentication where offered, and the Security Owner keeps a list of which applications support it.
Offboarding (Access Control Policy, 5.4)
With an identity provider: Offboarding and role changes. When People Operations confirms a departure, Engineering disables the Okta account, revokes sessions and tokens, removes membership of GitHub and AWS and Vercel, and rotates any shared or break-glass credentials the leaver could have known. The checklist is completed on the last working day and verified by the Security Owner within one business day. For internal moves, Engineering removes any access the new role's matrix entry does not require within five business days.
With none: Offboarding and role changes. When People Operations confirms a departure, Engineering disables the accounts on the master list, revokes sessions and tokens, removes membership of GitHub and AWS and Vercel, and rotates any shared or break-glass credentials the leaver could have known. The checklist is completed on the last working day and verified by the Security Owner within one business day. For internal moves, Engineering removes any access the new role's matrix entry does not require within five business days.
Approving a new application (Authentication and Password Policy, 5.7)
With an identity provider: New application onboarding. Before an application is approved under the Acceptable Use Policy, the Security Owner confirms whether it supports single sign-on and multi-factor authentication, integrates it with Okta where possible, and records any application that cannot meet this policy in the exception register with its compensating control.
With none: New application onboarding. Before an application is approved under the Acceptable Use Policy, the Security Owner confirms whether it supports single sign-on and multi-factor authentication, and records any application that cannot meet this policy in the exception register with its compensating control.
Three things to notice. First, the version without an identity provider does not pretend: it replaces the central account with a master list, and the leaver is removed from "the accounts on the master list". Second, it commits you to something. Item 4.3 of the Access Control Policy makes adopting an identity provider "a tracked security objective with a committed date". An auditor can ask for that date. If you do not plan to adopt one, edit the sentence before the policy is approved, because the written rule is what gets tested. Third, both versions check single sign-on support before a new application is approved, and send anything that cannot meet the policy to the exception register.
A working routine without SSO
- Keep the master list. One row per person per application: account, role, MFA on or off. A plain spreadsheet is enough. The site has no ready sheet for this list; the access review workbook is the record of each periodic review, one row per system, and the two go together.
- Turn MFA on in each application and keep the setting or the enrollment export. Start with the cloud console, source control and anything that holds customer data.
- Use one password manager. A unique password per application is what item 4.3 states when there is no identity provider.
- Offboard from the list, not from memory. The onboarding and offboarding checklist gives one dated line per application.
- Review access on a schedule. Export each application's users and compare them with the staff list. The free access review tool does the comparison in your browser and flags leavers who are still active.
- Record what cannot meet the rule. An application with no MFA, or SSO only on a plan you do not have, goes in the policy exception register with its compensating control and an end date.
This is more manual work each time someone joins or leaves, and it grows with every application. No fetched source gives a headcount at which a team should switch, and this page does not invent one. Linford's Hilary Stavrakas says the frequency of access reviews "should be determined by management based on the sensitivity and risk level of the system, as well as the volume and volatility of user accounts" (Linford). The same reasoning fits this choice: the more accounts change, the more a central switch is worth.
The other SSO question: customers signing in to your product
A buyer may ask whether your product supports SSO, meaning its employees sign in to your application with its own identity provider. That is a product feature and a sales requirement. It is a different question from how your own staff sign in to your own systems, which is what the sections above cover. The two get mixed up because the same vendors write about both: SSOJet's product, by its own description, sits in front of a SaaS application to accept customers' identity provider connections. If the question arrives in a security questionnaire, answer what you support today; the security questionnaire library has wording for access questions.
Frequently asked questions
- Does SOC 2 require SSO?
- Not by name, according to every source fetched for this page. SSOJet, which sells SSO, writes that "SOC 2 does not name single sign-on as a mandatory control". Secureframe lists single sign-on among practices that go beyond the baseline. Two articles by partners at the audit firm Linford & Co describe SSO as a benefit and an enhancement, and neither calls it required. What is tested is the outcome: who can sign in, how access is granted and removed, and whether your own written rules were followed.
- Is signing in with Google Workspace enough?
- It can be your identity provider. Linford's Isaac Clarke gives "the use of Microsoft or Google accounts to access various websites" as a common example of single sign-on, and the Policyseed generator lists Google Workspace next to Okta, Microsoft Entra and JumpCloud as an identity provider answer. What matters for the audit is what you can show from it: MFA enforced in the admin console, the list of applications that sign in through it, and a suspended account for each leaver. Applications that keep their own password stay on your list of exceptions.
- Do we need SCIM or automated deprovisioning?
- No fetched source says the criteria require it. SSOJet argues for it: it calls manual offboarding the place most teams fail and says SCIM "fixes the timing problem structurally". SSOJet sells SCIM directory sync, so read that as a seller's argument. The generated Access Control Policy asks for automated provisioning and deprovisioning only "where supported". Without it, the test is the same and the work is manual: every leaver removed from every application inside the window your policy states, with a dated record.
- What if a tool only offers SSO on a plan we do not pay for?
- Write it down as an exception with a compensating control, and keep the record. The generated Authentication and Password Policy has the Security Owner check single sign-on and MFA support before a new application is approved, and record any application that cannot meet the policy in the exception register. A compensating control can be a unique password from the password manager, the application's own MFA, and the application's accounts in the periodic access review. This page prints no price for SSO plans because none of the fetched sources gives one.
- A customer asked whether we support SSO. Is that the same question?
- No. That customer wants its employees to sign in to your product with its own identity provider. It is a product feature and a sales requirement, and a security questionnaire may ask about it. SOC 2 looks at how your own workforce reaches your systems. Much of what is written about SOC 2 and SSO, SSOJet's article included, is about the product feature, so check which of the two a source means.
- Can a five-person company get through SOC 2 without an identity provider?
- None of the fetched sources says it cannot, and none promises that it can; the CPA firm judges whether your controls are enough for your system. The work is a list: every application, every person's account in it, MFA switched on where offered, a per-application offboarding checklist and a periodic access review. Note what the generated policy commits you to when you answer that you have no identity provider: adopting one becomes a tracked security objective with a committed date. If you do not intend to adopt one, edit that sentence before you approve the policy.
Policyseed provides governance policy templates and AI tailoring. It is not legal advice and not a compliance guarantee. Management adopts the policies; the CPA firm performs the SOC 2 examination.