Other AI Tools

HokieAI covers most of what most people need. This page is the rest of what Virginia Tech has approved, and the specific jobs each one is actually better at.

Everything here is approved by VT. That matters more than it sounds: VT’s AI tools page states that “if an AI tool is not explicitly approved, it may only be used with low-risk, public data.” Anything not on VT’s list is a public-data-only tool, whatever its marketing says.

Google NotebookLM

Start here if you have a pile of sources and questions that cut across all of them.

NotebookLM is not a general chatbot. You give it documents and it answers only from those documents. VT lists it as available to “All VT students, faculty, and staff,” describes it as working with “user-uploaded sources (PDFs, Google Docs, URLs, videos, text files)” and supporting “multi-source analysis within a single workspace,” and rates it approved for high-risk data when you are signed in to your VT Google account.

Where it earns its place:

• Background reading before a meeting, so you can ask questions instead of reading all of it
• A fiscal year of policy updates in one notebook, so you can answer questions without digging through email
• Vendor proposals side by side, compared against each other rather than one at a time
• Getting up to speed on a project you inherited mid-stream

Try it: go to notebooklm.google.com, sign in with your VT Google account, create a notebook, and drag in two or three files you have been meaning to read. Then ask: “What do these agree on, and where do they disagree?” It answers from your sources, so when it says something surprising you can check it.

ARC — when the work has to stay on campus

ARC’s LLM service is Virginia Tech’s own. VT’s ARC documentation states it “offers on-prem Large Language Models (LLMs) for research, education, and administration” — administration included, not just researchers — and that “All Virginia Tech students, faculty, and staff may access the service at https://llm.arc.vt.edu. No separate ARC account is required to use the web interface.” There is no charge to individual users for the web interface.

Two things make it worth reaching for.

Nothing leaves the university. ARC’s documentation is unusually direct about this: “This service is hosted entirely on-premises within the ARC infrastructure. No data is sent to any third party outside of the university. All user interactions are logged and preserved in compliance with VT IT Security Office Data Protection policies.” When the question is not “is this tool approved” but “where does this data physically go,” ARC has the shortest answer of anything on this page.

You can try models you cannot get anywhere else at VT. ARC runs current open-weight models directly and swaps them as the field moves. ARC’s model list, read 28 July 2026, described them this way:

Model	ARC’s description
OpenAI gpt-oss-120b	”OpenAI’s flagship public model, best for fast responses.”
Z.AI GLM-5.2	”Top-performing model, best for complex tasks and coding.”
Moonshot AI Kimi-K2.6	”Top-performing model, best for complex tasks and multimodal. Use for image creation, edition, and analysis.”

That table is a snapshot of one day, and it will go stale — ARC states it “will add or remove models and scale instances dynamically to respond to user demand.” Check the model selector at llm.arc.vt.edu for what is actually running before you plan around a specific name. The churn is the point: if you want to know what a current open model is like, this is the sanctioned place to find out.

Per the same ARC documentation, “web search is enabled by default” and “image generation is disabled by default,” with image generation available through the interface’s integrations.

Try it: llm.arc.vt.edu, sign in with your VT credentials, and pick a model from the selector at the top of the page. Ask the same question of two different models and watch where they diverge.

Google AI Pro for Education

A paid add-on for people already living in Google Workspace. VT’s Software Service Center lists it as available to faculty and staff on annual renewal, “available as a purchasable add-on license beginning in March 2026,” and describes what it adds:

• “Access to advanced Gemini models (like Gemini 2.5) which provide better reasoning, summarization, and context generation”
• “Direct integrations with Google Workspace apps”
• “Increased limits on notebooks and sources per notebook in NotebookLM”
• “Deep contextual analysis, and audio/video overviews for study and research”

That third item is the one people miss — if NotebookLM has become part of how you work and you are hitting its ceilings, this is the upgrade that raises them.

The free Gemini app is already available to all VT students, faculty, and staff and is approved for high-risk data on a VT Google account. Try the free version first and see whether you actually hit a wall. VT publishes ordering details on the Google AI Pro for Education page; AAD IT can help you scope it.

Microsoft 365 Copilot

Worth knowing about, and worth being precise about, because two different things share the name.

Copilot Chat is included with VT M365 accounts, available to all students, faculty, and staff, and approved for high-risk data. It is a chat window. VT’s own description is blunt: “No Microsoft 365 app integration.”

Microsoft 365 Copilot is the paid add-on license that puts Copilot inside Word, Excel, PowerPoint, Outlook, and Teams. VT’s Software Service Center notes it “is an addon license and not the Copilot Chat service that is included with M365 A5 licenses,” and states that “as of December 2025, the new full-year cost of an M365 Copilot license is $216.” Licensing is annual and “expires August 31st each year,” and it “can be purchased mid-year at a prorated cost or renewed as part of your OUs EES (Enrollment for Education Solutions) bill.”

This is a department-level decision, not a personal subscription. It earns the money only if you genuinely spend your day inside Office apps. If most of your AI use is drafting and summarizing, HokieAI already does that for free.

Which one should I use?

Virginia Tech’s 4Help guidance puts it simply: “start with the tool that matches where your work already happens.”

• A general assistant, custom agents, most everyday work — HokieAI. VT describes it as the choice “when you want a flexible, institutionally managed AI workspace.”
• A stack of sources you need to question as a set — NotebookLM.
• Data that must not leave campus, or trying current open models — ARC.
• Work that lives in Word, Excel, Outlook, Teams — Microsoft 365 Copilot, if your department buys it.
• Work that lives in Docs, Sheets, Slides — Gemini, and Google AI Pro if you outgrow the free tier.

Not sure? HokieAI, NotebookLM, ARC, and the free Gemini and Copilot Chat tiers are all free to you. Start there.

Security and compliance

Virginia Tech classifies data by risk and approves each AI tool for a specific level. VT’s guidance is that “while several AI applications available at Virginia Tech are approved for use with high-risk data, the best practice is to reduce risk where possible and only use the lowest risk data that will still meet your needs.”

Approved-for-high-risk is a ceiling, not an instruction. The habit that keeps you out of trouble is using the least data that still gets the job done. VT’s guidance offers four questions to ask before the data goes in:

What level of risk would my data fall under? Is every data element necessary? Could I anonymize the data? Could I aggregate the data?

Most of the time the honest answer to the second question trims the job down to something low-risk on its own.

Where each tool currently stands, per VT’s own tools page. These are the published ratings, not a clearance to paste anything into anything — read the conditions below the table before you act on a row:

Tool	Availability	Data risk level
HokieAI	All students, faculty, and staff	Approved for high-risk data
ARC LLM gateway	All VT students, faculty, and staff	Approved for high-risk data — conditions apply, see below
Google NotebookLM	All VT students, faculty, and staff	Approved for high-risk data
Google Gemini App	All VT students, faculty, and staff	Approved for high-risk data
Google AI Pro	VT employees (license required)	Approved for high-risk data
Microsoft Copilot Chat	All VT students, faculty, and staff	Approved for high-risk data
Microsoft 365 Copilot (Licensed)	VT employees (license required)	Approved for high-risk data

Where the data goes differs even among tools with the same rating, and that difference is what to reason about:

• HokieAI — VT states that “inputs and outputs do not leave VT and are not used to train models.”
• ARC — “hosted entirely on-premises within the ARC infrastructure. No data is sent to any third party outside of the university.” Interactions are logged and retained under VT IT Security Office data protection policies.
• Google and Microsoft tools — VT describes these as having “enterprise-grade security and data protection when logged into VT Google account” or your VT Microsoft account. The protection is contractual and depends on you being signed in with your VT account. A personal account gets none of it.

Where the extra rules live. ARC’s high-risk approval is the one with visible conditions attached, and they matter. ARC’s documentation states it is “approved by VT Information Technology Security Office (ITSO) for processing sensitive or regulated data,” but in the same breath tells researchers to “consult with VT Privacy and Research Data Protection Program (PRDP) and the Office of Export and Secure Research Compliance regarding the storage and analysis of high-risk data,” and warns that “some high-risk data (e.g. data regulated by DFARS, ITAR, etc.) require additional protections and the LLM might not be approved for use with those data types.”

Read that carefully: on ARC, “approved for high-risk data” is not unconditional, and export-controlled data may not be approved at all. If your work touches ITAR, DFARS, or contractually restricted data, a rating on a web page — including this one — is not your answer. Ask before you paste.

And the outputs are still yours. Approval covers what goes in. Nothing covers what comes out: verify names, dates, numbers, quotations, and policy claims against the real source before anything leaves your hands.

Unsure where something falls? Ask AAD IT before the data goes in, not after.

Questions? Open a ticket or email aadithelp@vt.edu.

Last updated: July 2026