OpenAI ChatGPT Edu FAQ
OpenAI's ChatGPT Edu Agreement is tailored for the California State University (CSU) system, providing advanced AI capabilities for all employees and students. It offers the ability to build custom GPTs for sharing within campus workspaces, as well as the needed privacy and data protection, and other enterprise security such as single sign-on (SSO).
ChatGPT is an artificial intelligence (AI) chatbot technology that can process natural human language and generate a response, enhancing personal productivity and assisting with teaching and learning activities.
- Enrolled students (active)
- Staff
- Faculty
- All other affiliations are not eligible
Invitations to join the ChatGPT EDU Workspace are automatically sent based on your university affiliation.
- Students: You will receive your invitation to join the ChatGPT EDU Workspace within 24-48 hours of enrollment.
- Staff/Faculty: You will receive your invitation to join the ChatGPT EDU Workspace within 24-48 hours after HR updates PeopleSoft.
- Other affiliations (e.g., volunteers, persons of interest, alumni, emeritus) do not have access to ChatGPT.
While the CSU Chancellor's Office paid 16.9 million dollars for our access to ChatGPT, no funding came directly from CSUB to purchase this tool. This heavy investment in equity will help prepare our students the same as students from other, more well-funded CSUs.
We've been granted a special license to ChatGPT called EDU. This version of ChatGPT is something similar to ChatGTP Plus, valued at $20/month, but with no limits on 4o modal usage. We have access to most of the latest models, including 4.5, but 4.5 has a universitywide limit.
The CSU Chancellor’s Office has signed an 18-month (February 2025 - July 2026) contract for ChatGPT with an option to extend the agreement.
You can upload:
- Non-sensitive data that does not contain Level 1 (L1) classified information.
- Public or openly shareable data that does not require special access controls.
- De-identified data that has been stripped of personally identifiable information (PII) or other sensitive details.
- Non-copyrighted material that you have permission to share and process.
- For more details on data classification, refer to the CSU Data Classification guidelines.
Yes, the CSU ChatGPT Edu plan includes enterprise commercial data protection features like SAML Single Sign-On (SSO), domain verification, custom data retention windows, SOC 2 Type 2 compliance, and data encryption at rest and in transit.
Our agreement with OpenAI includes language that ensures your ChatGPT interactions and data are not used to train their underlying large language models or improve their services.
CSUB will not be monitoring individual user interactions or conversations in ChatGPT. CSUB will collect basic usage statistics to understand the extent of adoption and use.
The CSU contract explicitly prevents OpenAI from training its models on any data provided through our EDU accounts. This helps keep our data private and avoid the potential loss of intellectual property. Free and paid OpenAI accounts, as well as accounts with other Generative AI companies, do not have these protections. In fact, as a part of the terms you must agree to when creating an account, prompts, conversations, and files you create belong to the company and can be used to train their models.
Yes, the ChatGPT Edu plan allows institutions to build custom GPTs tailored to specific courses or administrative functions, which can be shared within the campus workspace to enhance learning and operational efficiency. Custom GPTs require the use of the GPT-4o model.
ChatGPT is just like any other CSU-provided tool. Like email, your data is not shared with or reviewed by anyone, except as a part of an officially sanctioned investigation. Your professors cannot see your conversation history or custom GPTs, even if they ask.
Yes, OpenAI provides applications for various platforms, including iOS, Android, Windows, and macOS, ensuring users can access ChatGPT across different devices.
- Free Plan Users: Faculty, staff, and students using the Free plan with their @csub.edu email address for their personal ChatGPT account will be prompted to archive or migrate their existing chat history into the CSU ChatGPT workspace upon logging in. This ensures a smooth transition to the new environment.
- Paid Plan Users: Faculty, staff, and students with Paid plans (e.g., Plus, Team, or Pro) who are using their @csub.edu email address for their personal ChatGPT account will have an opportunity to migrate their chat history before their paid account is deactivated once they are invited to ChatGPT EDU. Upon accepting their invitation to ChatGPT EDU, paid users will automatically receive a prorated refund for their remaining subscription period.
Student assistants should use their student ChatGPT Edu account for both student and student assistant activities.
Use of GenAI for assignments can range from absolutely forbidden to required. Your professor can specify if, when, and how AI should be used for a given course or assignment.
Academic integrity is core to quality education. This tool, like others, can be used unethically, so it's important to work with your faculty members to understand when and how it's permissible to use generative AI.
The ethical questions around Generative AI have a couple of angles, including the training data and the ethical use of AI. Models like the ones ChatGPT creates consume tons of literature from all over the internet which can contain the biases of the people or systems that created it. When using generative AI, we need to take care to evaluate the content it creates critically to avoid perpetuating biases, even unintentionally.
Training data can also contain copyrighted or protected intellectual property, which may not have been licensed to share, which raises concerns about the unethical use of that content. Considering the ethical use of AI in the higher education space, we should recognize that academic integrity is core to education. We need to make sure to use this tool ethically, only using it when approved by your professor and citing it when appropriate.
This new tool we've been given is like a calculator — when that was first introduced, people were concerned it would turn student's minds to mush. Quite to the contrary, after students learn how to do long division, they use the calculator to get to the next level — analysis and discussion of the meaning of the numbers.
Generative AI works by learning about things people have said in the past — consuming tons of information from the internet. As a result, the trained model makes its best guess based on what it's previously read to help answer the prompts it's presented with. However, when the model inferes the right response, it can make a mistake.
Training data can also be explicitly false — there may not be good examples for the prompt, or sometimes models infer responses inappropriately, which is also called hallucination. Making sure to provide well-developed prompts and to validate any information provided by generative AI can help you make sure you're using as accurate information as possible.
Self-paced learning resources are available:
- OpenAI, LLMs & ChatGPT (05:34)
- Multimodality Explained (10:24)
- Introduction to Prompt Engineering (05:52)
- Introduction to GPTs (06:40)
- ChatGPT Search (05:43)
- Advanced Prompt Engineering (08:50)
- ChatGPT for Data Analysis (04:48)
Energy usage comes in two flavors: training and usage. Training generative AI models like ChatGPT takes a boatload of energy up front — over 2-3 months, it's estimated that it takes 20-25 megawatts, which is the same as powering 20,000 American homes for the same time. Using Generative AI models is also called inference, and an excellent Epoch AI article has estimated that the energy of an average ChatGPT query is about the same as an average Google query.
In terms of environmental impact, two major topics are greenhouse gas emissions and water usage associated with training and inference. OpenAI does all of its training and inference on rented servers from Microsoft, which reported that its servers will be even better than carbon and water neutral by 2030, including the use of its servers for AI training and inference.