ChatGPT Edu adds enterprise-level security, enhanced privacy, and data agreements to ChatGPT, OpenAI’s generative AI assistant, ensuring that it can be leveraged responsibly with University data. No data entered in the Wharton ChatGPT Edu workspace is used for training OpenAI’s models.
ChatGPT Edu is offered alongside Wharton's Anthropic Claude Enterprise workspace — neither replaces the other. Consult the AI Tools and Resources guide for access requirements and the current list of Penn/Wharton-supported AI services and their restrictions.
ChatGPT Edu also features:
- Access to the latest OpenAI models, including GPT-5.6
- ChatGPT Work, OpenAI's agent mode for completing multi-step tasks and producing finished deliverables
- Codex, OpenAI's agentic coding tool
- Sharable custom GPTs within the workspace
- Image creation
ChatGPT Edu Benefits
Wharton ChatGPT Edu includes everything available in a free, personal account and adds several enhancements.
Enhanced Privacy
Data entered into the Wharton ChatGPT Edu workspace is never used by OpenAI to train its AI models. This means some University confidential data can be used with ChatGPT Edu. Please see the Security and Data section below for more information.
Models
ChatGPT Edu accounts have access to the latest OpenAI models, including GPT-5.6. You can switch to older models with a single click if those models are better suited to the task at hand.
Custom GPTs
GPTs are a custom-tailored version of ChatGPT that you can create with no code to accomplish various tasks. These GPTs can be shared with other users within the Wharton ChatGPT Edu workspace — see How to share GPTs within workspaces | OpenAI Help Center.
Custom GPTs created within the Wharton ChatGPT Edu workspace cannot be exported for access by ChatGPT users external to our workspace.
APIs
Wharton ChatGPT Edu includes API access for faculty and staff (students must work with a sponsoring faculty or staff member willing to both fund and accept risks for projects requiring API access). If you have a project that requires API access, please contact your Wharton Computing Representative. Refer to the ChatGPT Edu API Usage section below for API security information.
The New ChatGPT Desktop App
In July 2026, OpenAI consolidated its desktop offerings into a single unified ChatGPT desktop app for Mac and Windows. The former standalone Codex app and the traditional chat experience are now one application with three modes, accessible via a global switcher:
- Chat — quick questions and conversational help; the classic ChatGPT experience.
- Work — OpenAI's agent mode that completes multi-step tasks end to end, producing finished deliverables such as spreadsheets, slide decks, documents, and web apps. Work conversations sync across web, mobile, and desktop.
- Codex — a dedicated software development experience with inline diff editing, PR review, and multi-repo project support.
A few things to know:
- Existing Codex app installations are automatically updated to become the new ChatGPT desktop app.
- The previous ChatGPT desktop app has been renamed ChatGPT Classic.
- Developers who prefer the Codex experience can set Codex as the default view and choose the Codex icon for the app.
Downloading and Logging In
To download and install the ChatGPT desktop app, visit OpenAI's ChatGPT download page and follow their installation instructions for your platform.
To log into ChatGPT Edu with YourPennKey@upenn.edu on the web or in the desktop app:
- Go to chatgpt.com, or open the ChatGPT desktop app.
- Click Log in and enter your email address as
pennkey@upenn.edu, then click Continue. - You will be taken to the PennKey login screen. Enter your PennKey username and password, and confirm with Two-Step.
Verifying You Are Logged Into Wharton's ChatGPT Edu Workspace
Make sure you see the Penn shield displayed next to your name with "UPenn" underneath it at the bottom of the left sidebar to verify you are logged in with the correct account. You may have more than one ChatGPT account, so it is important to know which account you're currently logged into, especially if you're using University data.
Usage Limits
Not everything in ChatGPT Edu draws from your usage limits in the same way:
- Chat — traditional chat conversations do not draw from the usage limits described below.
- Work and Codex — these draw from a shared usage limit, metered over a rolling five-hour window. Larger and newer models draw down this limit faster.
Checking Your Usage
You can view your current usage in two places:
- In the ChatGPT desktop app, under your account Settings.
- Online at chatgpt.com/codex/cloud/settings/analytics.
ChatGPT Account Access
Wharton's ChatGPT Edu workspace is available to some members of the Wharton community. The following table lists who has access and how long that access lasts:
| Wharton Community Member Type | Length of Access |
|---|---|
| Standing Faculty & Full-time Staff | Contingent on an active affiliation with Wharton. |
| MBA, WEMBA, and PhD Students | From matriculation until graduation. |
| University students and TAs, while enrolled in Wharton graduate courses | For the duration of the enrolled course. |
Once eligible community members have an active PennKey account and the appropriate Penn affiliations, they will get an email invite directly from ChatGPT. Accounts inactive for six months may be removed. When you leave the University, access ends after a 14-day grace period; OpenAI retains the account for 30 days, so accounts restored within that window keep chats and history.
Security and Data
The University’s contract with OpenAI ensures that no data uploaded to Wharton’s ChatGPT Edu workspace is used to train any of their AI models.
Additionally, the contract includes data protections that enable ChatGPT Edu to be used with some kinds of data the University defines as confidential. Enterprise access does not automatically make every type of data acceptable — each service has its own approved data types, features, and restrictions. The following table lists Wharton data classifications and data types and if they can be used within ChatGPT Edu:
| Data Type | Allowed in ChatGPT Edu |
|---|---|
| Low | Yes. |
| Moderate - Not Involving PII or FERPA data | Yes - We recommend connecting with the Wharton Information Security Office on these usages of ChatGPT Edu. |
| Moderate - Involving PII or FERPA data | Yes, though you must engage with Wharton Information Security Office (ISO) for a risk review of the overall initiative’s data architecture and flow outside of the AI solution. |
| High* | No. Data with this classification should not be uploaded to ChatGPT Edu. Contact the Wharton Information Security Office if you require using data that falls into this category. |
| HIPAA/Protected Health Information | No. HIPAA data should not be uploaded to ChatGPT Edu. Contact the Wharton Information Security Office if you require using HIPAA data. |
Codex Security Considerations
Description: Codex is a code-focused agent accessed through the ChatGPT desktop app, CLI, or IDE extension. Features include code editing, inline diff editing, debugging, PR review, Git workflows, multi-repo projects, and terminal-based task execution in a user's local development environment.
Approved Data Tier: Up to Moderate (Non-PII / Non-FERPA)
Data Usage Notes: Code only. No secrets, passwords, API keys, PII, FERPA data, High data, sensitive institutional data, or non-code Moderate (Non-PII / Non-FERPA) data.
Codex is powerful: it can read files, edit work, inspect repositories, and run commands within the access you provide. Because these tools create risks beyond ordinary chat, review AI Coding Assistants at Wharton: Safe Use Guidance before using Codex. Moderate (PII and/or FERPA) data or anything above Moderate (Non-PII / Non-FERPA) requires ISO engagement and risk review. Interconnections with other systems, DUAs/contracts, compliance obligations, or data pulled from Moderate (Non-PII / Non-FERPA) systems also require ISO engagement, including review of how data is protected in transit.
Users must monitor agent actions and report information security or privacy concerns, unauthorized access, or suspicious behavior to security@wharton.upenn.edu.
ChatGPT Work Security Considerations
Description: ChatGPT Work is OpenAI's agent mode that completes multi-step tasks independently, gathering data from multiple sources and producing artifacts such as spreadsheets, slide decks, documents, and web apps. Work can connect to external tools and data sources via connectors.
Approved Data Tier: Up to Moderate (Non-PII / Non-FERPA)
Because Work acts autonomously across multiple steps and sources, users must review outputs and monitor agent actions. Connecting Work to any University-managed system or data source above the approved tier requires ISO engagement and risk review. Only Wharton-approved connectors may be used — see the Features section.
ChatGPT Edu API Usage
An interconnection is a direct integration or connection between an AI service and a system managed by the University. Connecting a generative AI service to a University-managed system or application, including through an API, requires an approved Penn or Wharton enterprise tool and review by the Wharton Information Security Office and Wharton Computing.
Managed system examples include (but are not limited to): Wharton Computing-managed computers, printers, and other endpoints; servers; on-premises and cloud applications; repositories and data sources; APIs; and IoT devices.
Sharing Chats or Custom GPTs
Sharing Custom GPTs and ChatGPT chats with specific users is disabled based on University Security Policy. To share with other users in the UPenn ChatGPT workspace, a Custom GPT must be shared to the full workspace via "Anyone at UPenn with the link."
See: Generative AI -- Best Practices and Resources for additional resources about using Generative AI tools at Penn.
Questions?
Please contact your Wharton Computing representative with any questions regarding ChatGPT Edu at Wharton.
Please contact Wharton’s Information Security Office with any questions you may have about data policies.