This page describes how Acquia AI processes data, specifies agent visibility, and details available assurance features in this release. Review this information with security teams before granting access to Acquia AI.

Model training

Acquia does not use customer data to train AI models.

Data visibility

Agent access depends on the active operational mode:

• Advisor mode: Accesses Acquia product documentation and basic account or subscription details. Advisor mode operates with user identity. Advisor mode cannot access sites, applications, or environments, and does not read log data.

This page describes how Acquia AI processes data, specifies agent visibility, and details available assurance features in this release. Review this information with security teams before granting access to Acquia AI.

Model training

Acquia does not use customer data to train AI models.

Data visibility

Agent access depends on the active operational mode:

• Advisor mode: Accesses Acquia product documentation and basic account or subscription details. Advisor mode operates with user identity. Advisor mode cannot access sites, applications, or environments, and does not read log data.

• Work mode: Accesses only assigned applications, sites, and context items added to projects. For delegated diagnosis tasks, the agent reads application or platform logs for specified environments and time windows, and reports identified patterns and evidence instead of raw log content.

For additional details, visit Modes in Acquia AI and Getting started with Acquia AI.

Credentials and sensitive values

The agent does not view credential values. Acquia AI removes passwords, keys, and contact details from records before agent processing.

Agent identity

In Work mode, the agent operates through a dedicated scoped identity rather than user permission sets. Consider these points during security reviews:

• The agent accesses only resources assigned by an administrator.

• The agent can access resources that specific users cannot reach directly, which enables users with Acquia AI access to interact with those resources through the agent. Assign only applications and sites that all users with access must affect.

In Advisor mode, this access scope does not apply because the agent operates with user identity and takes no action.

Outbound access

Acquia AI accesses external systems outside Acquia environments in three scenarios:

• Page governance checks
  Acquia AI reads specified pages through Acquia Web Governance to audit publicly accessible internet pages.

• Connected sites and applications
  Acquia AI retrieves logs from Acquia Cloud Platform for delegated diagnosis tasks.

• Slack review notifications
  When Slack integration is configured, notifications send one-way messages. External users cannot direct the agent or approve work from Slack, which prevents Slack workspaces from bypassing access controls.

In this release, only a user with an administrator role can enable or disable connectors, and settings apply to the entire organization. Include this behavior in security reviews.

Activity records

Acquia retains records of agent activity for internal operational purposes. This release provides no customer tools to review or export agent activity records.

Data residency

Acquia AI processes data exclusively in the United States East region in this release.

Compliance certifications

Acquia AI holds no formal compliance certifications in this release. ISO 27001 and SOC 2 Type 1 certifications are planned without announced completion dates. Contact the Acquia account team to verify current status before procurement reviews.

Available governance controls

Governance in Acquia AI relies on three core principles:

• Bounded capability set
  The agent cannot exceed functionality permitted in this release. The agent does not modify application code or infrastructure configurations.

• Scoped agent role
  The agent holds a dedicated Agent Team role, separate from human user roles. Roles consist of independently configured permissions. On Cloud Platform, an administrator can permit actions on one environment and withhold permissions on another.

• Two-stage human approval
  The agent proposes a plan and takes no action without approval. Publication requires a second, separate approval.

The following controls remain unavailable in this release:

• Per-user scoping of agent capabilities. Agent roles apply to entire projects and maintain uniform capabilities for all project users.

• Edits to Agent Team agent roles across products. On Source CMS, administrator controls remain limited to site selection.

• Customer-visible activity records.

• Automatic reversals for approved changes.

• Member restrictions for enabling or disabling connectors.

Agent roles and permissions

Review this section during security evaluations to define agent capability limits. The agent holds a dedicated role separate from human user roles. Roles consist of independently configured permissions:

• Cloud Platform
  Administrators configure exact permissions for the agent per environment. For example, administrators can permit cron task changes on non-production environments while withholding permissions on production.

• Source CMS
  The agent holds the Agent Team role built specifically for agent operations rather than human assignment, which prevents permission inheritance from users. Roles display for reference only, and Source CMS provides no permission controls to expand agent capabilities.

• Acquia DAM
  DAM administrators configure visibility scope directly inside Acquia DAM Admin Tools instead of Acquia AI. Agent permissions cannot exceed the permissions of that administrator.

The agent holds permissions exclusively on applications and sites assigned to its team. Resources outside assigned scope remain unavailable. For additional details, visit Governance and control management.