AI Services Terms

AI features are designed to reduce avoidable effort around customer work while preserving human responsibility. Features may support classification, transcription, summarisation, routing, suggestions, search, quality analysis, forecasting and content generation.

Customer must review AI output before relying on it, communicating it externally or taking action. AI may not be the sole basis for decisions producing legal or similarly significant effects unless the use is lawful, expressly documented and approved in writing by iconic.cx.

Output can be incomplete, incorrect, outdated, biased, unsafe or unsuitable. It is not legal, medical, financial, regulatory or other professional advice. Customer must test configurations for its context and maintain appropriate fallbacks and escalation.

Where AI communicates directly with an individual, Customer must give any legally required notice, including making AI involvement clear where it is not obvious. Customer must not remove a disclosure required by iconic.cx for legal or safety reasons.

We do not use identifiable Customer Data, prompts, communications, recordings, transcripts or output to train a shared or general-purpose model for unrelated customers unless an authorised Customer representative separately and expressly opts in in writing. Acceptance of the Customer Terms or use of an AI feature is not an opt-in.

Connected-service information is not used to train or fine-tune AI models. Any separate permission relating to other Customer Data excludes connected-service information and relevant derivatives. AI processing for a customer-authorised task is permitted only where it also complies with the source service’s rules. Providers processing the information for us are bound to the same permitted purpose and training restriction.

We may use identified subprocessors to deliver AI functionality. We require them to process Customer Data only for the permitted purpose and not for shared-model training without the same express opt-in. Applicable providers are identified through subprocessor information where required.

Customer retains its rights in AI input. To the extent permitted by law, Customer may use generated output, and we assign any assignable rights we hold in it subject to payment and the Agreement. Outputs may be similar across customers and are not guaranteed to be unique or free from third-party resemblance.

Customer must comply with the AI restrictions in the Acceptable Use Policy, including restrictions on unlawful discrimination, exploitation, fraud, unlawful surveillance, concealed AI involvement and unsupported high-risk use.

• Assign accountable owners and authorised administrators.
• Document purpose, data, review and escalation for material workflows.
• Keep meaningful human oversight for consequential decisions.
• Monitor output quality, complaints, incidents and changes in use.
• Disable or adjust features that cannot be used lawfully or safely.

AI inputs and outputs are Customer Data and follow the Agreement, DPA, customer settings and Data Retention and Deletion Policy. Temporary technical copies are limited to producing output, securing the feature, preventing abuse and investigating faults.

AI models, providers and capabilities may change. We will provide appropriate information about material purpose or control changes. We may suspend an AI feature where needed to address security, legal, safety or provider risk.