What is TulaneAI, and how is it different from ChatGPT?
| Quick Answer |
|---|
| TulaneAI is Tulane’s private AI platform. It gives you authenticated access to major AI models inside Tulane’s cloud environment, under Tulane’s security policies, without the LLM provider retaining or training on what you enter. |
What the Platform Is
TulaneAI is a no-code generative AI platform built for higher education by CloudForce in conjunction with Microsoft. It runs on Microsoft Azure AI Foundry and is hosted in Tulane’s own private cloud instance. What you see is a branded chat experience that behaves much like the consumer AI tools you already know, with one structural difference: the environment belongs to Tulane.
Why That Difference Matters
| Approach | What It Means | Data Control |
|---|---|---|
| Private cloud instance (TulaneAI) | Tulane-provisioned environment; models hosted by Microsoft; Tulane controls configuration | Governed by Tulane’s existing data security policies and safeguards |
| Third-party SaaS AI | Quick to start, limited customization | Less control over data policy |
| Public GenAI services | Any public service offering access to LLMs | No formal agreement with Tulane |
What You Can Do With It
Conversational chat, document and file work, and purpose-built agents are all supported. Capabilities vary by model — features such as web search, vision, and file libraries are not available on every model. A model that cannot use a particular platform feature is a limitation of that model, not a defect in the platform.
Related Resources
- AI Usage Guidelines Quick Reference Card: approved tools, data classification, governance tiers
- FAQ 2: naming — TulaneAI
- FAQ 3: data handling and retention
| Who This Applies To | Last Reviewed |
|---|---|
| All Tulane faculty and staff. Student access is a separate decision — see FAQ 5. | September 2026 |
Is nebulaONE the same thing as TulaneAI?
| Quick Answer |
|---|
| Yes. nebulaOne is the vendor platform. TulaneAI is what we call our instance of it. TulaneAI is the only name that should appear in university communications, training materials, and documentation. |
The Short Version
- nebulaOne — the CloudForce product name. You will see it in contracts, technical documentation, vendor release notes, and governance documents that reference the underlying platform.
- TulaneAI — the name of Tulane’s private instance. This is the name users see when they sign in, and the name to use in anything faculty or staff facing.
- Earlier working names — any placeholder name you encountered in pilot planning materials is retired. Do not carry it into new documents or presentations.
Why the Distinction is Worth Keeping
Using one consistent public name prevents the impression that Tulane is running several competing AI tools. It also keeps the vendor relationship legible in governance documents, where the platform is a component under a shared responsibility model rather than a Tulane-built system.
Related Resources
- AI CoE Communications Plan, naming and brand conventions
- FAQ 1: what the platform is
| Who This Applies To | Last Reviewed |
|---|---|
| Anyone writing or presenting about the platform, including CoE members, trainers, and department communicators. | September 2026 |
Does anything I type into TulaneAI get used to train AI models?
| Quick Answer |
|---|
| No. TulaneAI provides secure, authenticated access to AI models without data retention or training by the LLM providers. Your prompts and uploaded files stay inside Tulane’s cloud environment. |
Where Your Data Actually Lives
TulaneAI runs in Tulane’s own Azure environment. Conversations, configurations, and uploaded files are stored in Tulane-controlled Azure services and governed by Tulane’s existing data security policies. Model are run in Tulane’s cloud and return an answer; they do not keep the content or use it to improve their models.
What Is Retained, and For How Long
- Platform data is held in Tulane’s environment with standard backup and recovery windows configured at deployment.
- Usage is logged and tokens are tracked for administration, cost management, and auditability.
What This Does Not Mean
A private environment is not permission to enter anything you like. Data classification rules still apply in full: Level 4 (Restricted) data is prohibited without explicit review and approval.
Related Resources
- AI Usage Guidelines, data classification framework
- FAQ 5: what data is permitted in which tool
| Who This Applies To | Last Reviewed |
|---|---|
| All faculty and staff using TulaneAI. | September 2026 |
Who can use TulaneAI, and how do I get in?
| Quick Answer |
|---|
| Faculty and staff sign in at TulaneAI.Tulane.edu with their Tulane credentials through single sign-on. There is no separate account to request. Student access is a separate decision and is not currently part of the release. |
Getting Access
- Go to TulaneAI.Tulane.edu
- Sign in with your Tulane credentials. Authentication runs through Tulane single sign-on, so your identity and permissions follow your existing university account.
- If sign-in fails or you believe you should have access and do not, contact TulaneAI@tulane.edu.
On Student Access
Student access is a separate approval gate, not an oversight. It requires sign-off from the Office of General Counsel and the Provost’s AI committee before it can proceed. Faculty should not build course activities that assume student access to TulaneAI until that decision is announced.
Related Resources
- TulaneAI@tulane.edu for access and technical issues
| Who This Applies To | Last Reviewed |
|---|---|
| Faculty and staff. Students are out of scope for the current release. | September 2026 |
What information am I allowed to put into an AI tool?
| Quick Answer |
|---|
| Classify your data first, then choose the tool that matches it. Levels 1 and 2 are straightforward. Level 3 requires anonymization or a private-cloud model and approval. Level 4 is not permitted without explicit approval from the appropriate governance committee. |
The Four Data Levels
| Level | Examples | AI Tools Allowed |
|---|---|---|
| Level 1: Public | Published content, marketing, public website material | Any tool |
| Level 2: Internal | Financial data, FERPA records, employee records, donor information | Anonymized, or private-cloud models only with Governance approval |
| Level 3: Confidential | Financial data, FERPA records, employee records, donor information | Anonymized, or private-cloud models only with Governance approval |
| Level 4: Restricted | Social Security numbers, HIPAA/PHI, payment card data | Central IT-hosted models only, when available. Needs explicit review from Governance and OGC. |
How to Apply This in Practice
- Ask what the most sensitive element in your input is — not the average sensitivity. One FERPA record in an otherwise routine document makes the whole prompt Level 3.
- When you are unsure between two levels, treat it as the higher one and ask the CoE.
- Anonymizing means removing identifiers before the prompt, not asking the AI to ignore them.
Related Resources
- AI Usage Guidelines Quick Reference Card
- AI Usage Guidelines, full data classification section
- FAQ 6: approved and prohibited tools
| Who This Applies To | Last Reviewed |
|---|---|
| The classification framework is university-wide. | September 2026 |
Can I keep using ChatGPT, Claude or Gemini for Tulane work?
| Quick Answer |
|---|
| Not through personal or free accounts. Use the approved enterprise tools instead. The leading models are available to you inside TulaneAI, so this is a change in how you reach them, not a downgrade in what you can use. |
Tool Status
| Status | Tools |
|---|---|
| Approved | TulaneAI (nebulaOne), BoodleBox, Microsoft Copilot |
| Prohibited for Tulane business | Free or personally paid consumer LLM accounts, and any free LLM service. Access these models through TulaneAI instead. |
| Conditional — requires IT or CoE approval | Browser extensions, AI agents, and AI features embedded in existing SaaS platforms such as Oracle |
The Reasoning
A consumer account has no agreement with Tulane. There is no contractual control over how your input is retained, no audit trail, and no institutional recourse if something goes wrong. Consolidating onto governed platforms is itself a risk control — it converts invisible shadow AI into usage the university can see, support, and secure.
If a Tool You Need Is Not on the List
Submit a tool request to the CoE Lead. It is logged, forwarded to IT for technical and security assessment, and — if it passes — taken up at the next standing CoE meeting for governance review. Tools requiring budget go to the Executive Sponsor for final approval.
Related Resources
- AI Usage Guidelines, Section 9: AI Tool Procurement
| Who This Applies To | Last Reviewed |
|---|---|
| All COO Division staff, for any work performed on behalf of the university. | September 2026 |
What is the AI Center of Excellence, and what does it actually do?
| Quick Answer |
|---|
| The CoE is the operational AI governance body for the COO Division. It exists to clear the path for AI use — evaluating tools, approving assistants, and maintaining guidelines — not to slow requests down. |
Mission
The CoE exists to drive operational excellence, reduce costs, and mitigate risk by embedding disciplined, secure, and measurable AI so that Tulane’s COO Division delivers tangible results safely and at scale.
The Five Guiding Principles
- Human at the Helm. Every AI output requires human review before it is used, shared, or acted upon. No exceptions, at any tier.
- Guidelines Over Policies. Guidance that can be revised in days, rather than policy that takes months.
- Enabling Over Restricting. Governance builds highways, not roadblocks.
- Measurable Value. Every initiative traces to hours saved, dollars recovered, quality improved, or risk reduced.
- Internal First. Prove value inside the division before going public with it.
What It Governs, and What It Does Not
The CoE’s scope is the COO Division. It is expected to inform, and eventually integrate with, a university-wide AI Council. Until that Council exists, the CoE Charter is the governing document for AI within the COO’s operational footprint. The CoE reports directly to the Chief Operating Officer through the CoE Lead, with no intermediate governance layer.
Operating Rhythm
| Forum | Frequency | Who |
|---|---|---|
| CoE Steering Committee | Bi-weekly | All CoE members |
| Executive reporting dashboard | Monthly | CoE Lead to Executive Sponsor |
| AI model and tool newsletter | Monthly | Distributed division-wide |
| Comprehensive review | Quarterly | All stakeholders |
| Showcase / show-and-tell | Monthly or quarterly | Broader staff invited |
Related Resources
- CoE Charter, Sections 2 through 6
- RACI Matrix
| Who This Applies To | Last Reviewed |
|---|---|
| COO Division staff. Colleagues outside the division are welcome to use these materials, but the CoE’s formal authority stops at the division boundary. | September 2026 |
Do I need approval before I use AI for my work?
| Quick Answer |
|---|
| Often not. Personal productivity work with public or internal data needs no approval — start using an approved tool. Approval requirements begin when sensitive data, other departments, or automated decisions are involved. |
The Four Governance Tiers
| Tier | What It Covers | Approval |
|---|---|---|
| Tier 0 | Personal productivity, no sensitive data | None required |
| Tier 1 | Generative AI, some sensitive data, single department | Supervisor aware |
| Tier 2 | Cross-department, internal data, shared workflows | Supervisor aware; CoE Lead and department head approve |
| Tier 3 | Data-modifying, financial, privacy-affecting, or autonomous agents | Full CoE review |
If You Are at Tier 1 or Above
- Complete the Personal AI Assistant Submission Form, including your data classification and proposed tier.
- Submit it to the CoE.
- Your CoE reviews for data sensitivity and feasibility, targeting 24 hours.
- If approved, the CoE Lead confirms and you build with support.
- The finished assistant is documented and added to the repository so colleagues can reuse it.
Target turnaround is 72 hours from submission to decision.
The Ethics Checklist
Every initiative above Tier 0 is evaluated against the eleven-question ethics checklist. Most Personal AI Assistants clear it without difficulty — it is designed to catch edge cases. If your use case raises compliance complexity, material employee impact, institutional risk, or needs review by another body such as a Medical or Research committee, contact the CoE before proceeding.
Related Resources
- AI Usage Guidelines, Section 7: Ethics Checklist
- AI Usage Guidelines, Section 8: Personal AI Assistants and Scalable Solutions
- Personal AI Assistant Submission Form
| Who This Applies To | Last Reviewed |
|---|---|
| All COO Division staff building or using AI in their work. | September 2026 |
What do I do if something goes wrong?
| Quick Answer |
|---|
| Stop using the tool for that task immediately. Report it to the CoE Lead within 24 hours. Reporting quickly is the expected behavior, not an admission of fault. |
What Counts as an Incident
- Data exposure — sensitive information entered into an unapproved tool, or shared beyond its intended audience.
- Biased or discriminatory output that influenced a decision affecting people.
- Unauthorized tool use involving sensitive data, whether yours or someone else’s.
- Any situation where AI use may have caused harm to a person, a process, or the university.
The Steps
- Immediately: stop using the tool for the affected task. Do not attempt to delete or clean up first.
- Within 24 hours: report to your CoE Overseer and the CoE Lead.
- If the incident involves exposure of confidential or restricted data, also follow contact security@tulane.edu.
Preventing Most of Them
Human at the Helm is the single control that catches the most problems before they become incidents. Every AI output requires human review before it is used, shared, or acted upon. When AI supports a decision affecting people, budgets, or compliance, a qualified person owns that decision and its consequences.
Related Resources
- AI Usage Guidelines, Section 10: Incident Response
- CoE Charter, Section 3.1: Human at the Helm
- AI Safety and Operational Continuity Playbook
| Who This Applies To | Last Reviewed |
|---|---|
| All COO Division staff. Anyone may report an incident, whether or not they caused it. | September 2026 |