Before configuring an AI connector, decide which information your organisation permits each destination to receive. Then assign responsibility for credentials, sharing decisions and deployment approval. This makes the security review a practical decision process rather than a collection of broad assurances.
AIM’s Security & Governance page describes customer-controlled provider credentials, private work on the customer’s WordPress installation and explicit network-sharing scopes. These are useful starting points, but they do not answer every question about provider processing, retention or approval enforcement. The workflow below is a proposed approach for WordPress administrators and organisational buyers, not a statement that additional controls have been implemented or tested.
Start with permitted data destinations
Before entering credentials, identify the information a proposed workflow will use and where it may go. Separate these destinations rather than treating them as one AI service:
- The customer’s WordPress installation: where the AIM page says private working material remains.
- The connected AI provider: whose processing must be assessed separately from local storage and community sharing.
- AIM Network Hub: described as receiving data required for activation, directory and community functions, and deliberately shared learning.
- Public destinations: the AIM website, Directory or shared Library, where the page says approved material can be published.
A local working record does not, by itself, establish whether content is transmitted to a connected provider during an AI task. The supplied page does not document exact provider payloads, processing locations or retention arrangements. Treat those as open review questions, not as evidence that processing is either absent or unrestricted.
Build a responsibility matrix around evidence
A useful review record connects each material claim to an owner, supporting evidence and an unresolved question. Distinguish what the page states from what your organisation has independently verified.
Provider credentials and processing
The page says customers connect their own OpenAI Platform project through WordPress Settings → Connectors, and that the provider key is not sent to AIM Network Hub. Assign a connector administrator to verify project permissions and access arrangements. Separately, ask what information each intended workflow sends to the provider and review the applicable provider terms and settings.
WordPress access and private work
Assign an administrator to review hosting, WordPress roles, authentication and connector access. Record who can view working material, change configuration and authorise sharing. The statement that private material is not exposed through the community does not establish the installation’s complete access-control or retention arrangements.
Hub data and publication
Assign an appropriate organisational owner to review Hub data categories and sharing choices. Request a sufficiently specific account of activation and community data rather than treating broad category names as a complete inventory. For public material, document who may approve publication and how that approval is recorded.
Make the four sharing scopes concrete
AIM describes four scopes: Private, Network anonymised, Network attributed and Public. A practical review should explain the intended recipient, identity exposure and required approval for each choice.
Hypothetical example: choosing a scope for campaign learning
Suppose a marketing team wants to reuse lessons from a draft campaign. The following are illustrative decisions, not a report of completed work or verified product enforcement:
- Private: Keep the working draft out of community sharing. Review any connected-provider processing separately before using sensitive content in an AI task.
- Network anonymised: Propose sharing a general lesson without publicly identifying the organisation. Review whether distinctive wording, dates or context could still reveal its identity; the scope label alone does not demonstrate effective anonymisation.
- Network attributed: Share the lesson with the organisation’s identity attached only after the appropriate owner approves that disclosure.
- Public: Prepare a version specifically approved for publication, checking that client details, internal plans and confidential material are excluded.
The key question is not simply “Which label sounds safest?” It is “Does this recipient, identity exposure and content match our authorised use?”
Use an owner-assigned deployment checklist
Bring configuration and governance checks into one review record. This can begin as a document; a dedicated readiness tool would be a separate proposal requiring feasibility checks.
For each item, record an owner, status, verification note, unresolved risk and approval decision. Cover hosting, roles, authentication, provider permissions, connector access, sharing defaults, retention, moderation, approval flows and internal policy. Link to authorised evidence rather than copying credentials or sensitive working content into the checklist.
Define which unresolved issues prevent rollout. For example, if nobody can establish whether a planned workflow sends restricted information to a provider, that workflow should remain unapproved until the question is resolved. A completed checklist is a decision record, not a security certification.
Test understanding before relying on the guidance
The proposed review format should be tested, not assumed to prevent mistakes. Ask representative administrators and buyers to complete five scenario tasks using the current guidance and a revised version. Include distinguishing local storage from provider processing, choosing sharing scopes and identifying the publication approver.
Record accuracy, destination misunderstandings and completion time. A proposed initial threshold is four of five readers answering correctly, with no critical destination misunderstandings. This is a small usability check—not statistical proof or evidence of security effectiveness. Current comprehension and checklist-completion baselines have not been supplied.
Keep procurement evidence separate from compliance claims
A dated responsibility matrix can also support supplier-review conversations. Mark statements as documented, independently verified or unresolved, and retain the limits alongside them.
AIM’s page explicitly cautions that separate credentials do not themselves establish compliance with ISO standards, SOC requirements or UK data-protection law. The practical next step is therefore to document permitted destinations, resolve material processing questions and obtain accountable approval—not to turn architectural descriptions into compliance assurances.

Add an example, conclusion or question.
Short, specific responses help AIM Community improve future articles and product decisions.