AI vendor due diligence

What AI or models power this system, and how are they updated?

Model and provider chain, material changes and how changes are communicated.

What data do you collect and what is it used for?

Data categories, purpose, inputs, outputs, telemetry and secondary uses.

Do you use our data to train or improve your models?

Whether organisational data trains or improves models, and whether it is isolated from other customers.

Are your recommendations informed by other customers’ data?

Whether pricing or other recommendations draw on non-public data from other customers, including competitors, and how it is segregated.

Where is our data stored and processed, and does it cross borders?

Locations, jurisdictions, subprocessors and transfer safeguards.

How long do you retain our data, and can it be deleted? How do you carry out applicable deletion requests and restrictions on data use or sharing across connected systems, subprocessors and shared-property databases?

Retention for inputs, outputs, logs and backups; deletion process and evidence; how applicable deletion requests and restrictions on data use or sharing are carried through connected systems, subprocessors and shared-property databases; any exceptions or limitations and how fulfilment is confirmed.

Who can access our data? Where AI agents access our systems, does each agent have a distinct identity, limited permissions and an exportable record of its actions, and can its access be revoked independently?

Vendor personnel, subprocessors, support arrangements and access controls; distinct agent identities, permissions limited to the approved role, exportable action records sufficient for investigation, and independent revocation of agent access. Action records remain subject to appropriate privacy and retention controls.

What security controls and assurance evidence support this service?

Authentication and access controls, including multi-factor authentication (MFA) for human accounts; relevant security testing and independent assessments; any certifications and their scope, validity and applicability to the service being purchased. Certifications should support, not replace, assessment of the proposed use.

What can the AI do without human approval?

Action boundaries, permissions, human review, override and suspension capability.

How can we disable or suspend the AI feature, who can do so, and what happens to the underlying service?

The available shutdown or suspension control; who has authority and access to use it, including property-level access where relevant; whether vendor intervention is required; the effect on the underlying service; and the fallback arrangements.

How will we know if something changes or goes wrong?

Material-change and incident notification, relevant records and escalation.

What happens if your service, underlying model provider or cloud provider becomes unavailable?

Which hotel functions stop or degrade; available manual or alternative processes; outage communication, recovery arrangements and how interrupted transactions are reconciled.

What happens when the relationship ends?

Data return or deletion, records, continuity and exit assistance.

Who owns the outputs this system generates for us, and do you indemnify us against third-party claims arising from them?

Ownership or licence of outputs, any vendor claims over outputs, and the scope, limits and exclusions of indemnity and liability for third-party claims.