Your research data is yours. Your work product is yours.
University procurement asks two questions about an AI platform that institutional marketing pages rarely answer plainly: who owns what the platform helps us produce, and where does our research data go. ArthurAI™ University Learning Edition answers both by construction. The institution is the data controller and the owner of the work product. Eve-Education™ reasoning is trained on synthetic data — never on institutional data — and research data stays inside the institution's tenant boundary by architecture, not by promise.
Who owns the work product
The institution owns the curricula, assessments, advising records, and scholarly artifacts produced in its deployment — including the ones a faculty member produces with AI assistance. The faculty member is the faculty of record and the authoring authority for that work; the platform is an instrument they direct, reviewed and attested before anything reaches a student, a grade, or a publication.
ArthurAI™ claims no intellectual-property interest in the instructional materials, assessments, or research artifacts the platform helps a faculty member produce. In routine deployment, Eve-Education, LLC is a data processor acting on the institution's documented instructions — not a co-author, not a rights-holder, not a licensor of the institution's output back to itself.
When research is genuinely joint
There is one exception, and it is always explicit. Where MindHYVE.ai and an institution choose to co-author formal research — as in the joint AI-competency work in PharmD education with California Northstate University — authorship, attribution, and intellectual-property terms are governed by a separate written research agreement negotiated before the work begins. Co-authorship is an opt-in collaboration the institution elects, never a byproduct of using the product. Absent such an agreement, the default posture stands: the institution owns its work product and ArthurAI™ is a processor.
No institutional data is ever used for training
Eve-Education™'s reasoning capability is built on Eve-Genesis™ — a proprietary synthetic reasoning corpus generated to capture high-quality pedagogical reasoning, calibrated by experienced reviewers. It is synthetic by construction. No student data, no faculty work, no research data, and no institutional content enters the training pipeline at any point. The architecture cannot leak what is never present.
This is the procurement-decisive distinction for a research institution: the platform does not improve its models by reading your grant work, your unpublished findings, or your students' submissions. Frontier-model providers composed into the reasoner receive only the content required to serve a given request, under contract terms that prohibit training on that content.
More on the synthetic-data posture: Eve-Genesis · data handling.
Where research data lives
Research data is held inside the institution's tenant boundary, enforced at five independent layers — database queries filter by institution identifier, file storage is tenant-scoped, the cache keyspace is partitioned by tenant, the API binds every request to the authenticated institution, and the frontend carries institution scope from the session. A cross-tenant data path requires bypassing all five simultaneously. Research data does not cross institutional boundaries by construction.
Eve-Grid™ deploys with residency awareness: North American institutions in U.S. Azure regions, European institutions in EU regions under Standard Contractual Clauses for any necessary transfer, and African institutions in the closest applicable Azure geography aligned with local data-protection regimes. An institution's data residency requirement is satisfied at the substrate, not patched at the application layer.
Architecture detail: five-layer multi-tenant isolation.
Processing roles
For institutional deployments, the institution is the data controller for personal data and the holder of education records under FERPA. Eve-Education, LLC is the data processor under Article 28 GDPR (and analogous state-law constructs) and the school official under FERPA, processing data only on the institution's documented instructions. Research data supplied to or generated within the platform inherits the same controller/processor structure — the institution directs the purpose and means; ArthurAI™ executes within them.
Human-subjects and sensitive research data
ArthurAI™ is a teaching-and-learning platform, not an IRB-governed research instrument. Where an institution conducts research involving human subjects, the institution's IRB and research-compliance office remain the governing authority; the platform does not assume that role. Reasoning telemetry the platform collects to operate — error rates, latency, queue depth, workflow completion — never includes the content of educational records or research data. If a deployment requires the platform to process regulated research data, that scope is defined in the institutional agreement and the data-handling addendum, not assumed by default.
What the platform never does with your data
- It never uses institutional data — student, faculty, or research — to train a model.
- It never moves research data across an institution's tenant boundary.
- It never asserts authorship or an IP interest in the institution's work product absent a separate, signed research agreement.
- It never sends research data to a frontier provider under terms that permit that provider to train on it.
- It never relocates data outside the contracted residency region without the institution's documented instruction.
What this means for procurement
A university procurement and counsel team can attach this posture to an RFP response without caveat: the institution remains the controller and the owner of its work product; training is synthetic-only and contractually walled from institutional data; isolation and residency are architectural, not configuration flags; and the one path to joint IP — co-authored research — is an explicit, negotiated agreement, never a default of using the product. Every claim here maps to a control documented in the Trust Center.
See also: compliance posture · security + audit · disclosures · university deployments.