UNCG Spartan AI Opens With L4 Data Approval for ChatGPT, Claude

UNCG's Spartan AI Puts ChatGPT, Claude, Grok and Mistral Behind One Azure-Hosted Front Door​

The student newspaper The Carolinian reported the rollout on September 22. Its details line up with an announcement from UNCG's Information Technology Services (ITS) posted September 14. ITS calls Spartan AI "a new secure AI workspace that brings multiple AI models together in one university-supported environment." It says the service is "powered by nebulaONE" and lists models from OpenAI, Anthropic, xAI and Mistral AI. The Carolinian's account named Grok, Claude and ChatGPT but left out Mistral.

According to ITS, the workspace can be used for:

• Drafting and revising content.
• Summarizing and analyzing information.
• Brainstorming and exploring topics.
• Working with documents.
• Organizing related work in Projects.
• Building specialized Agents for repeated tasks or workflows.

Projects and Agents matter most here. With them, Spartan AI becomes a place to set up reusable workflows, and it can do more than a chat window with a model picker.

The university makes three main security commitments. Prompts and data entered in Spartan AI are not used to train AI models. The service works with information a user can already access in Microsoft 365, including L3 and L4 data. And it runs inside UNCG's Microsoft Azure cloud environment. The Carolinian wrote that Spartan AI "will not be used for the training of open sources models." The official wording is broader: no training of AI models at all, with no limit to open-source ones.

Getting started takes four steps: go to the Spartan AI site, sign in with a UNCG account, choose a model and start a conversation. ITS suggests that newcomers begin by asking the model how it could help with their work or studies.

nebulaONE Is the Azure Gateway Doing the Work Under Spartan AI​

Spartan AI is UNCG's brand on a commercial platform. nebulaONE comes from Cloudforce, a Maryland company. Cloudforce's launch material describes it as a multi-modal Generative AI platform, built exclusively on Azure. The vendor also says it offers turn-key integration with all the top AI models from OpenAI, Anthropic, Meta, Mistral and more, deployed with a single click.

Microsoft has promoted the product itself. A Microsoft Education blog post calls Cloudforce a Microsoft Supplier of the Year in 2024. It says Cloudforce built nebulaONE on Azure to use its built-in security and privacy features and is working with dozens of higher-education institutions. The same post frames the pitch around a governance problem: when students and staff adopt their own unsecured AI tools, it creates concerns with IT governance, security, privacy, and data protection, and it limits the ability to scale AI throughout the institution.

Other deployments show the usual pattern. The University of Oxford's AI Competency Centre announced a nebulaONE pilot in late 2025. It described the platform as a gateway to a wide range of large language models, including those from OpenAI and open-source models, hosted on the University's private Microsoft Azure tenancy. The PR Newswire launch release cited UCLA Anderson School of Management, which chose nebulaONE for its fully managed AI platform, seamlessly deployed in Azure with enterprise-grade security and privacy.

UNCG's statement that Spartan AI "runs within UNCG's Microsoft Azure cloud environment" fits that model: the gateway sits in the institution's own tenancy and doesn't send users to each vendor's public consumer app. UNCG hasn't said how each model is hosted or reached behind the gateway. Grok and Claude, for example, may be connected differently from OpenAI's models. Campus IT staff evaluating similar products should ask that question directly.

Cloudforce's pricing claim also matters for IT budgets. The company says customers pay only for what you consume, with built-in budgets and chargeback reporting, avoiding traditional per-user licensing fees. UNCG hasn't published what Spartan AI costs or how it's billed. Consumption pricing would still explain why a university can open a multi-model tool to its whole population during a pilot without buying seat licences for everyone. That link is an inference; UNCG hasn't confirmed it.

Spartan AI's L3 and L4 Approval Sets It Apart From Free Copilot and Public Chatbots​

For people at UNCG, the most useful detail is where Spartan AI sits in the university's data-classification rules. The Carolinian summarized UNCG's levels this way. L3 is confidential, moderate-risk data such as employee reviews and FERPA-protected student records. L4 is restricted, high-risk data such as Social Security numbers and protected health information.

UNCG's "Permissible Use Cases" guidance on its AI hub lists which tools are approved for which levels:

The practical point is this. Claude and ChatGPT are approved only up to Level 2 when used through their individual paid plans. Reached through Spartan AI, they fall under an approval that extends to Level 4. The same split applies on the Microsoft side: the Copilot built into Windows 11 is limited to Levels 1–2, while the enterprise Microsoft 365 Copilot is approved through Level 4. The contract, hosting and privacy terms behind the product decide what data you can use, and the model name doesn't.

UNCG's guidance explains the basis for these approvals. Any large language model or AI chatbot the university buys must include explicit privacy guarantees that prompts, conversations and interactions are not used to train or improve the underlying models. AI-enabled products must also pass a UNCG ITS Pre-Purchase Software and Hardware Review, with a full Security Posture and Risk Assessment when applicable, repeated on an ongoing basis. Spartan AI's "not used to train AI models" promise is therefore a condition the university sets for approval.

UNCG's Own Rules Are Stricter Than the Spartan AI L4 Approval​

Approval for L4 data doesn't mean any sensitive record can go into any prompt. The Spartan AI announcement tells users to "remove identifying details when they aren't needed," to ask when unsure about particular information, and to check AI-generated content for accuracy before using or sharing it.

The wider guidance goes further in several places:

• For data analysis, UNCG says to remove all identifying information from a dataset before uploading it, even when the AI system offers privacy protection. ITS 6-TECH provides guidance on de-identification.
• For summaries, it warns that AI output may be incomplete or inaccurate and strongly encourages checking the primary documents before any critical university decision.
• Users should never share a Banner ID, username or password with any AI tool.
• Some uses are called high risk or outright inappropriate for any AI tool: evaluating admissions applications, screening job applicants' résumés and evaluating competitive bids. Grading student work, personnel or disciplinary decisions and legal analysis require prior consultation with the responsible unit head and other campus offices.

These are general UNCG rules for AI, and Spartan AI doesn't enforce them technically. A staff member could build a Spartan AI Agent to summarize job applications. University policy says they shouldn't, whatever the tool's data classification.

For coursework, instructors decide. ITS says AI use in a course is governed by the instructor's policy and may be prohibited, and all coursework remains subject to UNCG's Academic Integrity Policy. The Carolinian notes that using AI in a class that forbids it may count as an academic integrity violation. The university's list of prohibited uses starts with completing academic work in a way the instructor doesn't allow.

The Carolinian's Microsoft 365 and Copilot Claims Go Beyond UNCG's Announcement​

The Carolinian says Spartan AI is "integrated within Microsoft 360." That should read Microsoft 365. The paper also says the safeguards include protections "within Microsoft 360 Copilot." The official announcement is narrower. It says Spartan AI can be used with information a user already has access to in Microsoft 365, and it doesn't describe Spartan AI as part of Microsoft 365 Copilot or built into the Microsoft 365 apps. The permissions-based wording suggests Spartan AI follows each user's existing Microsoft 365 access and doesn't grant new access. That reading is an inference. UNCG hasn't published the technical mechanism.

The Carolinian also reports device and account limits: access only for certain UNCG accounts on devices that meet university security requirements, with data kept on UNCG-managed resources. The ITS announcement says only "sign in with your UNCG account" and doesn't list device requirements. Keeping data on UNCG-managed resources is consistent with the Azure hosting statement. Anyone who hits a sign-in block on a personal laptop should contact 6-TECH, which can explain the actual conditional-access rules better than either published account.

A 12-Month Spartan AI Pilot Backed by UNCG's AI Oversight Committee​

UNCG calls Spartan AI part of a 12-month AI evaluation program meant to give the university "real-world experience with different AI technologies." The aim is to learn how people use the tools, where they add value and what the community needs, and then use that to guide longer-term decisions about AI services. The announcement doesn't give an end date, and it doesn't commit to keeping Spartan AI, or nebulaONE, after the pilot.

That fits the multi-model design. When students and staff can pick among OpenAI, Anthropic, xAI and Mistral models in one place, the university gets evidence on which models people actually choose for which jobs. It also avoids committing to one vendor's campus-wide licence before it has that evidence. UNCG is asking for feedback through a form about what works, what's confusing and where AI could help.

Training started with the launch. The on-demand course "Spartan AI: Connected and Secure" covers starting and managing chats, choosing a model for a task, writing prompts with UNCG's PACE method, organizing work in Projects and using Agents for repeated tasks. ITS has scheduled live webinars for Monday, October 5, at noon and Wednesday, October 14, at 10 a.m. The Carolinian describes the live sessions as in-person, but ITS lists them as webinars, so confirm the format when registering. ITS news also lists an "AI for Work, Teaching, and Productivity" pilot showcase on Tuesday, October 20.

The governance work behind this goes back more than a year. UNCG's Chancellor's Council commissioned the Artificial Intelligence Oversight Committee in February 2025. It is chaired by Casey Forrest, Associate Vice Chancellor for Enterprise Technology Infrastructure and Chief Technology Officer, and has a representative from each campus division. Its charge covers strategic guidance for AI in learning, academic integrity, professional development and operational efficiency. Its goals include policy updates, training, ethical business use of AI and gathering community input to guide AI investments. In April 2025, UNCG said the committee's first major deliverable was a central AI hub with campus guidelines and a list of ITS-approved tools. That hub now carries the classification table above. The published record doesn't say whether the committee itself chose nebulaONE.

What this means for you​

If you're a UNCG student, faculty member or staff member, Spartan AI is the approved place to use Claude, ChatGPT, Grok or Mistral with university information. It's approved for data your individual consumer or Business/Pro subscriptions aren't. Students should check each course's AI policy before using it for classwork. Staff handling HR, admissions, procurement or grading should read the high-risk list before building anything. IT staff at other institutions should look at UNCG's model: an Azure-tenant gateway, approval tied to a published classification table, and a time-limited pilot with a feedback loop.

• Spartan AI has been available to all current UNCG students, faculty and staff since September 14, 2026. You sign in with your UNCG account.
• UNCG has approved Spartan AI for Level 1 through Level 4 data. Free AI tools are limited to Level 1, and Windows 11's built-in Copilot and Copilot Chat to Levels 1–2.
• Even inside Spartan AI, remove identifying details you don't need, never enter a Banner ID or password, and check outputs against the primary documents.
• Don't use any AI tool, Spartan AI included, for admissions review, résumé screening or bid evaluation. Get prior consultation before using it for grading or personnel decisions.
• The on-demand "Spartan AI: Connected and Secure" course is available now. Live webinars run October 5 at noon and October 14 at 10 a.m.
• The service is a 12-month pilot with no published end date. Use the feedback form to affect whether and how it continues.

Spartan AI shows a university treating generative AI as governed infrastructure. The models sit in its own Azure tenant, the approvals are tied to a published data-classification table, and the usage rules are stricter than the technical approval. Over the next twelve months, the pilot will produce the usage data UNCG says it will use to decide whether Spartan AI, or some other arrangement, becomes a permanent campus service. The feedback students and staff submit during the pilot is part of that record.