Microsoft Copilot, (formerly known as Bing Chat), a generative artificial intelligence (AI) tool that can be used to generate or analyze text and images, is now available for use at Princeton University. Copilot is currently the only generative AI tool made available by the Office of Information Technology (OIT).
Data security and privacy
Copilot is available in two modes: “Copilot for consumers” and “Copilot with Enterprise Data Protection.” When you are logged in to Copilot with your Princeton University account, you are using Copilot with Enterprise Data Protection which better protects information.
In order to help secure Princeton data and reduce risk to the University, log in to Copilot with your Princeton University account, to enable the following protections:
- (a) your prompts and responses are not used to train the underlying large language models, and
- (b) chat data sent to and from Copilot is encrypted.
It is recommended that you avoid writing prompts that include restricted or confidential information, including proprietary research data.
What is Copilot and how is it different from Copilot for 365?
Copilot provides access to generative AI functionality built on the multimodal large language model GPT-4, the text-to-image model DALL-E 3, and search results from Microsoft Bing.
Copilot for 365, which integrates AI into Outlook, Word, Excel and other Office applications is currently being evaluated.
How to use Copilot
To use Copilot, users must sign in to Copilot with their princeton.edu account using their netID and password.
1. On the Copilot website, in the top-right corner, select Sign in, then Sign in with a work or school account. Follow the prompts for netID login and Duo two-factor authentication.
Generative AI and Princeton’s data classification
When handling University information, it is critical to store, share, and access it safely and consistent with the University’s data classification guide.
If you plan to use Copilot without your Princeton.edu account or another Generative AI tool that has not been provided by OIT, please ensure that the information you submit is neither restricted (e.g. social security or credit card numbers), confidential (e.g. personnel information), or unrestricted within Princeton (e.g. web-based resources for University use or University ID numbers).
This submitted information can be used to train the underlying model and potentially become available to users outside of Princeton University.
Generative AI and data accuracy
Regardless of the tool or mode, AI responses can be biased, inaccurate, inappropriate, or may contain unauthorized copyrighted information. Always review, validate, and iterate on responses provided by any Generative AI tool before relying on them or sharing them broadly.