AI & Private LLM Acknowledgment
AI-Assisted Functions and Private Model Processing Acknowledgment · Supplement to the Terms of Service and the Business Associate Agreement
This Acknowledgment sets out the terms under which the artificial-intelligence features of the Nuvanei Health platform (the "AI Functions") are made available to a health care practice (the "Practice") and its Authorized Users. It describes what the AI Functions do, the actions they are not permitted to take, where and how they process information, the controls that govern them, and the responsibilities of the individuals who use them. Each Authorized User accepts this Acknowledgment before using the Services. Capitalized terms not defined here have the meanings given in the Terms of Service.
Article 1. Definitions
1.1 "AI Function" means a feature of the Services that uses a machine-learning model to generate a draft, summary, classification, suggestion or other output for review by an Authorized User.
1.2 "Approval Queue" means the component of the Services in which every output of an AI Function is held until an Authorized Reviewer approves, edits or rejects it.
1.3 "Authorized Reviewer" means an Authorized User whose role permits them to approve a given category of output, as configured by the Practice and enforced by the Services (for example, only a credentialed prescriber may approve a drafted refill request).
1.4 "Private Model Deployment" means an instance of a large language model or other machine-learning model that is provisioned for Nuvanei Health's exclusive use within Nuvanei Health's cloud environment, that does not share capacity or data with other customers of the model provider, and that is operated under a Business Associate Agreement with the hosting provider.
1.5 "Model Provider" means the vendor that supplies and hosts a model used in a Private Model Deployment.
Article 2. Description of the AI Functions
2.1 The AI Functions are assistive. They prepare material for human review and do not act on their own. The AI Functions available at the Effective Date are:
| AI Function | Inputs | Output placed in the Approval Queue | Authorized Reviewer |
|---|---|---|---|
| Intake assistant | Intake forms, history and prior notes within the Practice's tenant | Pre-visit summary, identified gaps and suggested questions | Provider or clinical staff |
| Inbox assistant | Patient messages and conversation history | Triage category, routing suggestion and a draft reply | Role assigned by the Practice; clinical replies require a provider |
| Laboratory assistant | Laboratory results, the Practice's configured ranges and prior results | Organized result summary, flagged changes and a draft plain-language explanation | Provider |
| Protocol assistant | Active prescriptions, supplement protocols and refill history | Draft refill or renewal request; consistency notes on protocols | Credentialed prescriber |
| Billing assistant | Invoices, membership status and payment history | Draft payment reminders and prepared superbills | Billing administrator or owner |
2.2 Nuvanei Health may add, modify or retire AI Functions. Any new AI Function will be subject to this Acknowledgment and described in the Documentation before it is enabled for a Practice.
2.3 The Practice may disable any AI Function for its tenant at any time through the Services' settings. Disabling an AI Function does not affect data already recorded in the audit log.
Article 3. Human Review and Prohibited Capabilities
3.1 Approval Queue. Every output of an AI Function is placed in the Approval Queue. No output is communicated to a patient, entered into the clinical record as an interpretation, submitted as an order or prescription, transmitted to a Partner Service, or applied to a charge, invoice or payment request until an Authorized Reviewer has reviewed it and recorded an approval. An approval is attributed to the reviewing individual, not to the AI Function.
3.2 Prohibited capabilities. The AI Functions do not, and the Services do not provide them the technical means to:
- render or record a diagnosis;
- prescribe, renew, modify or discontinue a medication, or transmit a prescription to any pharmacy or electronic-prescribing network;
- order, cancel or modify laboratory or imaging tests, or transmit an order to any laboratory;
- create, change or discontinue a treatment plan;
- communicate with a patient without an approved action by an Authorized Reviewer;
- charge a payment method, issue a refund or send a payment request;
- make determinations regarding eligibility for care, coverage, employment or any other decision that produces legal or similarly significant effects on an individual; or
- access data belonging to any tenant other than the Practice in which the request originated.
Electronic prescribing is performed exclusively by a credentialed prescriber acting under their own identity through a certified third-party partner.
3.3 Patient transparency. Where a message to a patient was drafted with the assistance of an AI Function, the Services record that fact. The Practice may configure the patient application to display a notice that the message was prepared with assistance and reviewed by the care team, and the Practice is responsible for complying with any disclosure requirements applicable to it.
Article 4. Model Processing and Data Handling
4.1 Private Model Deployments. Clinical, patient-related and practice text submitted to an AI Function is processed only by Private Model Deployments. Such text is not transmitted to public or consumer artificial-intelligence services.
4.2 Data residency. Private Model Deployments used for the Services are located in the United States.
4.3 No training. Neither Nuvanei Health nor any Model Provider uses Practice Data, patient data, prompts or outputs to train, fine-tune, evaluate or otherwise improve any model for use outside the Practice's tenant. Model Providers are contractually prohibited from retaining prompts and outputs beyond the processing of the request, other than transient operational logs required for abuse monitoring, which are retained for a defined period not exceeding thirty (30) days and then deleted.
4.4 Minimum necessary. Each AI Function is designed to include in a prompt only the information reasonably necessary to perform its task, consistent with the minimum-necessary standard under 45 C.F.R. § 164.502(b). Tenant isolation is enforced at the database layer, so a request cannot retrieve data from another Practice.
4.5 Notifications. Push notifications, text messages and emails generated by the Services never contain Protected Health Information, regardless of whether an AI Function was involved in preparing the underlying content.
4.6 Subcontractors. Model Providers are subcontractors of Nuvanei Health under the Business Associate Agreement and are bound by written agreements imposing the same restrictions and conditions that apply to Nuvanei Health. The current list of Model Providers is available to the Practice on request.
Article 5. Governance, Logging and Accountability
5.1 Attribution records. For each output of an AI Function, the Services record the identifier and version of the model used, the time of generation, references to the inputs used, and the identity of the Authorized Reviewer who approved, edited or rejected the output, together with the time of that action. These records form part of the Practice's audit log and are available to the Practice for review and export.
5.2 Change management. Nuvanei Health evaluates AI Functions before release and before any change of underlying model, using test data that contains no real patient information, and maintains records of those evaluations. Nuvanei Health will give the Practice notice through the Services of any change to an AI Function that materially alters its behavior.
5.3 Monitoring. Nuvanei Health monitors the AI Functions for errors, unsafe outputs and misuse. The Practice and its Authorized Users may report suspected errors through the Services; Nuvanei Health will investigate reports and correct confirmed defects within a commercially reasonable time.
5.4 Incident reporting. Any unauthorized use or disclosure of Protected Health Information involving an AI Function or Model Provider is reported under the incident and breach provisions of the Business Associate Agreement.
Article 6. Limitations and Disclaimers
6.1 Outputs of the AI Functions may be incomplete, outdated, inaccurate or inconsistent, and may reflect limitations or biases of the underlying model or of the information available in the Practice's tenant. Outputs are drafts prepared for professional review; they are not medical, legal, financial or billing advice and do not constitute a standard of care.
6.2 Interpretations that reference "optimal" or functional ranges configured by the Practice reflect the Practice's clinical preferences rather than laboratory reference ranges or established clinical guidelines. The Services label such interpretations accordingly when they are presented to patients.
6.3 The AI Functions do not replace, and are not intended to replace, the professional judgment of licensed practitioners, and are not intended for use in emergencies or for time-critical decisions.
6.4 THE AI FUNCTIONS ARE PROVIDED SUBJECT TO THE WARRANTY DISCLAIMERS AND LIMITATIONS OF LIABILITY IN THE TERMS OF SERVICE. APP NAME MAKES NO WARRANTY REGARDING THE ACCURACY, COMPLETENESS OR CLINICAL SUITABILITY OF ANY AI-GENERATED CONTENT.
Article 7. Responsibilities of the Practice and Authorized Users
7.1 Practice responsibilities. The Practice will (a) configure Authorized Reviewer roles so that only appropriately credentialed individuals can approve clinical, prescribing and billing outputs; (b) maintain policies governing its use of AI-assisted tools and patient transparency consistent with applicable law and professional standards; (c) ensure that its Authorized Users accept and comply with this Acknowledgment; and (d) not attempt to circumvent, disable or alter the Approval Queue or the attribution records described in Article 5.
7.2 Authorized User responsibilities. Each Authorized User will (a) review every AI-generated output for accuracy, completeness and appropriateness before approving it, signing it or relying on it; (b) edit or reject outputs that are inaccurate, incomplete or inappropriate; (c) approve only outputs within the scope of their license, role and competence; (d) not submit information to an AI Function beyond what is necessary for the task; (e) report suspected errors, unsafe outputs or misuse to the Practice owner and to Nuvanei Health; and (f) comply with the Practice's policies regarding the use of AI-assisted tools and patient disclosure.
Article 8. Acknowledgment
By accepting this Acknowledgment, the Authorized User confirms that they have read and understood Articles 1 through 7; that they understand that the AI Functions generate drafts for professional review and do not diagnose, prescribe, order, transmit or communicate with patients on their own; that clinical, prescribing and billing decisions remain the sole responsibility of the licensed professionals at the Practice; and that they will discharge the responsibilities described in Article 7. Nuvanei Health records the version of this Acknowledgment accepted, the identity of the accepting user, the date and time of acceptance, the network address and a cryptographic fingerprint of the accepted text.
This document is a pre-release draft provided for review and has not yet been approved by legal counsel. A final version will be published and, where required, re-acceptance requested.