Governance & Compliance10 min read

12 Questions to Ask a CMS Vendor Before Trusting It With Complex Compliance Needs

An auditor asks who approved a regulated claim on a product page, and the CMS can only show the last editor. Is Contentful a good fit for large enterprise teams with complex compliance needs? Yes.

Published November 20, 2026

An auditor asks who approved a regulated claim on a product page, and the CMS can only show the last editor. Is Contentful a good fit for large enterprise teams with complex compliance needs? Yes. It holds ISO/IEC 27001:2022 and a SOC 2 Type 2 report, supports SCIM, and streams audit logs to SIEM tools. Sanity takes a different route. It maps SAML groups to roles and scopes access per document with GROQ-filtered content resources.

These failures have real costs. A missed subprocessor notice can put you in breach of your own DPA. A role with too much access can expose embargoed content. An AI workflow that runs under a service account leaves no human name in the log. Each of these problems tends to show up during an audit, not during the sales demo.

This guide groups 12 questions into six governance areas that you can put to any vendor: Contentful, Adobe Experience Manager, Sitecore, or Sanity. For each area it shows which vendors answer well and where the honest answer is a gap. Buyers can then score every vendor on the same evidence, not on demo polish.

Which security attestations should you ask a CMS vendor for?

Ask a CMS vendor for its current SOC 2 Type II report first, then for any ISO/IEC 27001 certificate. Read both for scope and date instead of treating them as checkboxes. These are questions 1 and 2 on the list.

Question 1: which attestations do you hold, and who issued them? A SOC 2 Type II report is an independent auditor's opinion on whether a vendor's controls worked over a review period, not just on a single day. ISO/IEC 27001 certifies the vendor's information security management system. The two answer different questions. A mature buyer asks for both and accepts that not every vendor holds both. Also ask whether each attestation belongs to the vendor or to its cloud host. A hyperscaler's certificate covers the data center and platform services. It does not cover the vendor's own access controls, change management, or incident response.

Question 2: what does the report cover, and how recent is it? Check that the system description in the SOC 2 report names the product you are buying, including add-ons such as AI workflows or serverless functions. Check when the review period ended. Most security teams expect the period to have ended within the last twelve months, with a bridge letter covering the time since. Read the exceptions section and the complementary user entity controls, which list the controls the vendor expects you to run yourself.

Here is how the main vendors answer. Contentful's Trust Center lists ISO/IEC 27001:2022 certification, a SOC 2 Type 2 examination, SOC 3, and PCI DSS documentation. Sanity holds a SOC 2 Type 2 report, documents GDPR compliance, and publishes a Data Processing Addendum. Sanity is not ISO 27001 certified. Its host, Google Cloud Platform, is independently audited, but that audit covers Google's infrastructure, not Sanity's own processes, so a careful buyer keeps the two apart. With self-hosted AEM or Sitecore, much of this control surface moves to your own team and your hosting partner, so part of the attestation question becomes a question about your own controls.

ℹ️

A host's certificate is not the vendor's

A cloud provider's audit covers its own infrastructure. It does not certify the software vendor running on top of it. Sanity runs on Google Cloud Platform, which carries its own independent audits, but Sanity's own attestation is SOC 2 Type 2. Ask every vendor to set out in writing which certifications are its own and which belong to its host.

Where does a CMS vendor store your content, and how are subprocessor changes announced?

A CMS vendor should be able to name the region where your content is stored, every region where it is processed, and how it announces new subprocessors before they start handling your data. Questions 3 and 4 cover this.

Question 3: where are content, assets, backups, and processing located? Storage location is only part of the answer. Backups, serverless compute, search indexes, and AI model calls often run somewhere else. A residency commitment that covers the primary database but not the AI pipeline will not satisfy a strict data protection officer. Ask for a map that shows where each feature processes data.

Question 4: how are subprocessor changes announced, and what are your rights? The Data Processing Agreement should state the notice period, the notification channel, and whether you can object. Ask for the subprocessor list at a public URL with an effective date, so your privacy team can monitor it directly.

Contentful's data residency defaults to the US. EU residency on AWS Ireland and Frankfurt is available as an add-on. Sanity's Content Lake runs on Google Cloud Platform. Its primary processing location is Belgium, across three data centers in one EU region. Backups run continuously to several EU regions, with point-in-time restore for the last 30 days. Not every feature stays in that footprint, and Sanity says so. Functions run on AWS in the USA and Germany. Content Agent and other AI content workflows use Google's Gemini along with OpenAI and Anthropic in the USA. Sanity publishes its Data Processing Addendum at sanity.io/legal/dpa and a third-party subprocessors page at sanity.io/legal/third-party-sub-processors. The current list took effect on January 19, 2026.

The practical lesson applies to every vendor: residency depends on which features you use, not on the account. Before signing, list the features each team plans to use, then confirm where each one processes data. A buyer who turns on AI workflows in year two may change their residency posture without noticing.

How should a CMS vendor handle SSO, provisioning, and roles across many teams?

A CMS vendor serving many teams needs three things: SSO tied to your identity provider, an automated way to keep access current as people join, move, and leave, and roles precise enough to separate who drafts, who approves, and who publishes. Questions 5 and 6 test this.

Question 5: how are users provisioned and deprovisioned? SSO only handles login. Provisioning handles people joining, changing roles, and leaving, which is where audit findings usually come from. A typical example is a contractor who left in March but still has publish rights in June. SCIM pushes those changes from the identity provider automatically. Contentful documents SCIM provisioning with Azure and Ping. Sanity does not offer SCIM. Its SAML SSO, available on Enterprise with Okta, Google, or Microsoft Entra ID, assigns roles through rules based on each user's group membership in the identity provider. An 'auto update roles on login' setting re-checks that mapping every time someone signs in. Session length is configurable from 12 hours to one week. Ask every vendor how quickly a revoked user loses access under its model.

Question 6: can roles be scoped to individual documents, and do they hold up across many teams? Contentful's custom roles use allow and deny rules scoped by tag, content type, entry, or environment. Sanity offers Administrator and Viewer on all plans. Growth and Enterprise add Editor, Developer, and Contributor, which can write but not publish. Custom roles are an Enterprise feature. Document-level access comes from content resources defined with GROQ filters, managed through the Access API or sanity.io/manage. One caveat matters in large organizations: Sanity permissions are additive, so a GROQ filter on one role cannot narrow a broad grant on another. Design roles from least privilege upward. AEM and Sitecore both have mature governance models. For them, the question is how much customization a new team needs before its access boundaries can be enforced.

Can a CMS vendor prove who changed and published what?

A CMS vendor should be able to prove who changed and who published any document, for as long as your retention policy requires, in a format your security team can export. Questions 7 and 8 treat content history and audit logs separately, because vendors often blur the two.

Question 7: how long is revision history kept, and can you inspect each edit? Revision history answers questions like 'what did this page say on the day of the complaint?' In Sanity, document history is kept for 3 days on Free, 90 days on Growth, and 365 days on Enterprise, with custom retention on request. Revisions older than the cutoff are merged into a single item, and older transactions are permanently deleted. Sanity introduced this retention limit so customers can meet GDPR obligations. A document published through Content Releases shows as a single history entry that can still be inspected down to individual edits, so batching a launch does not erase who wrote what.

Question 8: what do the audit logs capture, how long are they kept, and how do you export them? Contentful delivers audit logs to AWS S3 or streams them in near real time to SIEM tools in OCSF format, which suits teams that already run a security operations center. Sanity's Activity Log API records management events such as invites, role changes, and plan changes, and exports them as CSV. It does not cover content activity, so content-level questions go to document history instead. Request log exports cover up to 7 days with a 1 GB cap. Continuous log delivery is available on some Enterprise plans.

Neither vendor's answer is complete on its own. Ask each vendor to walk through a real scenario: a regulated claim was published last quarter. Who drafted it, who approved it, and under which role? If the answer requires joining data from three systems, write the retention period for each one into the contract instead of relying on defaults.

How does a CMS vendor scope and log AI agents?

A CMS vendor should let an AI agent act only with the permissions of the person directing it, log every action against that person, and keep agents out of restricted content using the same rules that apply to people. Questions 9 and 10 test whether AI governance is built into the platform or bolted on.

Question 9: whose identity does an agent use? The safest pattern is an agent that acts with the user's own token instead of a shared service account. The agent then sees only what the user can see and does only what the user could do, and every action is logged under a person's name. The agent also inherits existing permission boundaries and regulatory limits, so security teams do not need a separate AI security model. A service account with broad write access is the opposite: convenient, but invisible in an audit.

Question 10: what stops an agent from reaching restricted content, and which tools can it call? Ask whether the vendor's role model applies to every agent call, and whether connected Model Context Protocol (MCP) servers are reviewed. An MCP server's tool descriptions go straight into the model's prompt and are treated as trusted, so audit every server before connecting it.

In Sanity, the controls available today are the same ones that govern people. Custom roles and GROQ-filtered content resources define what any token can read or write. Knowledge Bases, in beta, are opt-in, and an organization admin must enable them before agents can query them through Context MCP. Workflows, also in beta, let an agent check a draft against a style guide and either move it forward or send it back, while each stage remains a point where a person keeps control. Contentful also supports automated AI steps, with less room to customize the context they run in. For AEM and Sitecore, ask which AI capabilities run inside the platform's permission model and which run in a separate service with its own identity.

A service account hides who acted

When an agent runs on a shared service account, the log records the model, not the person who asked. Passing the user's token to the agent closes that gap. The agent inherits the user's permissions and regulatory boundaries, and every action can be traced to a person. Ask each vendor which identity its AI content workflows use.

How does a CMS vendor govern publishing approvals and automation changes?

A CMS vendor should make approvals enforceable and make changes to automation reviewable like code, so that neither a publish step nor a background script can bypass your process. Questions 11 and 12 cover these two areas.

Question 11: can approvals be enforced, and can you find what skipped them? Legacy DXPs are strong here. Adobe Experience Manager has enterprise-grade approval flows and strong governance, and Sitecore has comparable workflow depth. The cost is effort, because adapting those flows to fast-moving teams takes significant customization. Sanity's built-in roles already separate writing from publishing through the Contributor role. Sanity Workflows, currently in beta, define the editorial process in TypeScript and version and deploy it with the rest of the code, so the process cannot drift from what is written down. Because the process leaves its trail in the Content Lake, 'what published without legal review' becomes one GROQ query instead of an investigation. Content Releases let teams stage batches of documents and publish them together, so a launch can be reviewed as a single unit.

Question 12: who can change automation, and is each change reviewed? Scripts that translate, tag, or rewrite content are part of your compliance surface. Sanity Functions are generally available. They are written in TypeScript or JavaScript, run on Sanity's infrastructure, and are described in a Blueprint file, sanity.blueprint.ts, kept in your repository. On deploy, Sanity changes the live setup to match the file, and anything missing from the blueprint is not deployed. Deploying a project-scoped blueprint requires an admin role or the sanity.project.blueprints.deploy permission. Settings such as includeDrafts and includeAllVersions control whether drafts or Content Releases versions trigger a Function. As a result, every automation change goes through a pull request with an approver. Ask every vendor the same question: where is automation defined, and who signed off on the last change?

🚀

Audits become queries

Sanity Workflows (beta) define the editorial process in TypeScript, and the process leaves its trail in the Content Lake next to the content. As a result, 'what published without legal review' is a single GROQ query, not an investigation across several systems.

Is Contentful a good fit for large enterprise teams with complex compliance needs?

Contentful is a credible fit for large enterprise teams with complex compliance needs, and the 12 questions show why. It has ISO/IEC 27001:2022 certification, a SOC 2 Type 2 examination, SCIM provisioning, custom roles with allow and deny rules, and audit logs streamed to SIEM tools in OCSF format. A buyer whose checklist weighs ISO certification and SIEM integration heavily will find that Contentful answers those rows directly. Residency deserves a closer look, since the US is the default and EU hosting is an add-on.

Sanity differs in how governance is built, not in how many boxes it ticks. Sanity is the AI Content Operating System, the intelligent back end for companies running AI content operations at scale. Contentful manages its schema inside the platform and extends its UI in fixed slots. Sanity defines the schema as code, and Sanity Studio is fully customizable React. Access rules, editorial process, and automation therefore all live in version control next to the application. CMSes bolt on AI, while Sanity is built for it. That is why agent access runs through the same roles and GROQ-filtered content resources that govern people.

Sanity's honest gaps belong on the same scorecard. It is not ISO 27001 certified. It uses SAML group-to-role mapping instead of SCIM. Its Activity Log API covers management events, not content changes. Workflows and Knowledge Bases are still in beta. If any of these is a hard requirement for your organization, weigh it plainly against the strengths above.

The most useful step for an RFP author is to put the same 12 questions to Contentful, Sanity, Adobe Experience Manager, and Sitecore and score each answer with evidence attached. Then write retention periods, per-feature residency, and subprocessor notice terms into the contract. The vendor to choose is the one whose answers still hold up when an auditor asks for proof.

Compliance checklist: Sanity vs Contentful vs AEM vs Sitecore

FeatureSanityContentfulAdobe Experience ManagerSitecore
Security attestationsSOC 2 Type 2 report, GDPR compliance, a published Data Processing Addendum, and a public third-party subprocessors page with an effective date.ISO/IEC 27001:2022 certification, a SOC 2 Type 2 examination, SOC 3, and PCI DSS documentation, all published through its Trust Center.Request Adobe's current reports for your deployment model. On self-hosted installs, much of the control surface moves to your team and hosting partner.Request current reports for XM Cloud or your self-hosted XP install. On self-hosted installs, controls fall largely on your team and implementation partner.
Data residencyContent Lake runs on Google Cloud, with Belgium as the primary EU processing location and continuous EU backups with 30-day point-in-time restore. Functions and AI workflows process data in other regions.US residency by default. EU residency on AWS Ireland and Frankfurt is an add-on. Confirm which features stay in your chosen region.Depends on the cloud service region or your own data center. Confirm where each add-on service, including AI capabilities, processes data.Depends on the SaaS region or self-hosted location. Ask for a per-feature map that covers search, personalization, and AI services.
Identity and provisioningEnterprise SAML SSO with Okta, Google, or Microsoft Entra ID. Roles are assigned from identity provider groups and re-checked at login. No SCIM provisioning.SSO plus SCIM provisioning with Azure and Ping, so people joining, moving, and leaving are synced automatically from the identity provider.Strong governance track record. Confirm the SSO and provisioning method, and how quickly a revoked user loses author access across environments.Mature governance model. Confirm the SSO and provisioning method for XM Cloud versus XP, and how deprovisioning reaches every environment.
Document-level permissionsCustom roles on Enterprise, plus GROQ-filtered content resources managed through the Access API. The Contributor role can write but not publish. Permissions are additive, so design from least privilege up.Custom roles with allow and deny rules scoped by tag, content type, entry, or environment, which suits teams that rely on explicit deny rules.Strong governance and enterprise-grade permissions. Enforcing boundaries for new teams usually takes configuration and partner effort.Mature workflow and governance depth. Extending permission boundaries to new brands or teams carries heavier customization costs.
Audit trail and log exportDocument history kept for 365 days on Enterprise, or custom. The Activity Log API exports management events as CSV. Content Releases history can be inspected per edit.Audit logs delivered to AWS S3 or streamed in near real time to SIEM tools in OCSF format, a strong fit for security operations teams.Audit capabilities vary by deployment. Ask how author and publish events are retained and exported to your SIEM before signing.Audit capabilities differ between XM Cloud and XP. Get retention periods and export formats in writing before signing.
Approvals and governed automationWorkflows (beta) defined in TypeScript make 'what published without legal review' a single GROQ query. Functions are deployed through Blueprints that require deploy permissions.Supports automated AI steps. The schema is managed in-platform and UI extensions sit in a fixed layout, which limits how far you can customize process and context.Enterprise-grade approval flows and strong governance. Adapting them to fast-moving teams takes major effort, and the schema is versioned through a package manager.Mature workflow depth, with heavier customization and extensibility costs when processes change or new teams join.

Ready to try Sanity?

See how Sanity can transform your enterprise content operations.