CMS Remediation Agent Security and Compliance

Summary

This article explains access, credential handling, and customer responsibilities when using the CMS Remediation Agent. For setup instructions, supported CMS requirements, and the remediation workflow, see the CMS Remediation Agent overview.

Access and approval

The Use CMS Remediation permission controls who can start CMS analysis and approve a proposed change. The connected CMS account determines which content the agent can access and change. Review proposed content and the CMS destination before approving an update.

Manage the Siteimprove role and the CMS account separately. Use a dedicated integration account with only the permissions needed for the content in scope. Review its access when responsibilities or the connection’s scope change.

Credentials and secure connections

CMS connections require HTTPS. Siteimprove stores CMS credentials encrypted and does not display them after saving. You can edit or delete a connection in Siteimprove. Revoke the credential or block the dedicated account in the CMS when access is no longer needed.

Keep credentials out of page content, shared documents, and remediation instructions. Follow your organisation’s policies for personal or confidential information and AI-assisted services.

Publication and verification

Approving an update to published WordPress content can make the change live immediately. In Drupal, check the resulting publication status against your site’s moderation workflow. Review the CMS’s revision or content history where available and keep your editorial and recovery processes in place.

After the public page reflects an update, re-check it in Siteimprove. A successful CMS write does not by itself establish accessibility conformance or legal compliance.

How AI is used

AI Remediate generates a code suggestion for a selected accessibility issue. The CMS Remediation Agent then investigates the connected CMS and prepares a proposed change for review. Applying that change requires a separate approval.

AI provider

The agent runs on Amazon Bedrock. The CMS remediation workflow uses AI to help investigate the issue and prepare a proposal. AI output should be reviewed in the context of the page and the intended content.

Information used in the workflow

When you select Generate in AI Remediate, the selected HTML snippet is sent with a prompt to produce a suggested fix. Analyze in CMS is an additional step: the agent can retrieve relevant CMS content, URLs, field information, and metadata to locate the affected material and prepare a proposal.

Treat the CMS account’s read permissions as part of your information-handling review. Depending on those permissions, the account can expose content beyond the public page, including material that has not been published. Limit its access to the intended scope.

Model training

Inputs to and outputs are not used by AWS or third-party model providers to train Amazon Nova, Amazon Titan, or third-party models.