This Privacy Statement describes the ways in which Harvard University may access and use personal information about individuals in connection with providing the AI Sandbox application managed by Harvard University Information Technology (HUIT). In this Privacy Statement, “AI Sandbox” refers to the HUIT-managed generative AI chatbot accessible by authorized users at sandbox.ai.huit.harvard.edu. 

This Privacy Statement applies to all users of the AI Sandbox service, regardless of their relationship to Harvard. 

The personal information about users that may be accessed by Harvard in connection with providing the AI Sandbox includes name, email address, affiliation, and user chat history. 

We view log data about individuals (whenever possible, separate from an actual person’s name) only as permitted by, and subject to the requirements and restrictions of, the University Policy on Access to Electronic Information.  

The personal information that can be compiled by Harvard in connection with providing the AI Sandbox is used for internal learning and usage analytics.  

Harvard may review and share internally within the University users’ IP addresses, role type, School or unit, and the activities they perform using the AI Sandbox to administer the service, provide technical support, and help diagnose problems. Harvard may use and retain email or other communications to staff in order to process inquiries, respond to requests, and improve services.  

Other uses of personal information by Harvard are described in other parts of this Privacy Statement.  

If the user discloses personal information in the context of a query submitted to the AI Sandbox, the AI Sandbox will, consistently with its ordinary function, share that information with Large Language Model (LLM) provider(s) designated by the user. 

Each such LLM provider may have its own privacy policy. Harvard is not responsible for the privacy practices of such third-party resources or for the content they generate. Please see below (“Information Protection”) for more information about assurances Harvard has obtained from LLM providers. 

Other sharing of personal information by Harvard is described in other parts of this Privacy Statement.  

In addition, Harvard may share and disclose personal information, including outside of Harvard, if Harvard believes that doing so is necessary or appropriate to: satisfy any applicable law, regulation, legal process or governmental request; investigate compliance with or enforce this Privacy Statement or any acceptable use policy for the AI Sandbox; detect, prevent or otherwise address fraud, security or technical issues; or protect, defend or enforce the operations, property, rights and safety of users, Harvard, Harvard affiliates or others. We notify users as soon as possible when we take an action that impacts their specific account or system. 

Harvard may use cookies and related technologies (“Cookies”) in the AI Sandbox application to gather information when users navigate through the AI Sandbox, in order to enhance and personalize the experience, to help provide services available through the AI Sandbox, to analyze service usage, and to help improve the service. To find out more about cookies, visit https://www.internetcookies.com.

You can review your Internet browser settings, typically under the sections “Help” or “Internet Options,” to exercise choices you have for certain Cookies, or you can opt out of the collection and use of some Cookies through tools like the Network Advertising Initiative opt-out page. If you disable or delete certain Cookies in your settings, you may not be able to use features of our websites. 

There are security measures in place to help protect against the loss, misuse, and alteration of personal information under Harvard’s control. No data transmission, processing, sharing or storage by any party can be guaranteed to be completely secure. 

The AI Sandbox offers a single interface that enables access to Large Language Models (LLM) provided by OpenAI, Amazon Web Services, Google, and Microsoft. 

• Harvard has enterprise agreements in place with all LLM providers to protect Harvard data and ensure certain privacy protections are in place for users of the AI Sandbox.
• Per their published service descriptions and agreements with Harvard, LLM providers do not store any user input or generated responses.
• Per their published service descriptions and agreements with Harvard, LLM providers do not utilize any user input to train their models.
• All connections between the user’s browser and the AI Sandbox, and from AI Sandbox to the models, are encrypted. 
• Chat sessions and associated data are stored in a HUIT-managed database that is protected by enterprise IT policies and can be used to store data up to Harvard data classification level 3.
• Chat sessions are private by default and can only be shared through direct user interaction. Even when shared, your name and any messages you add after sharing stay private.

Please contact ithelp@harvard.edu with any questions or concerns about this Privacy Statement.  

If you are located outside the United States, please click here for additional Harvard privacy disclosures that may apply to you. 

This Privacy Statement may be amended from time to time. Any such changes will be posted on this page. This Privacy Statement was last updated on September 30, 2024.