¾«¶«AV

Generative AI Guidelines for Security and Privacy

What is generative AI?

Generative AI refers to a class of artificial intelligence models that can produce content, such as text, images, code, or music, in response to user prompts. This is distinct from other forms of machine learning used for analysis, classification, or prediction. Examples include large language models (LLMs) like those used in chatbots and text-to-image generators. You can learn more about the technology from the Center for Teaching and Learning (CTL).

How generative AI tools are being used

Many applications now include AI features like chatbots, summarization, content generation, and suggestions. As these technologies proliferate, new uses are identified regularly. Here are  several examples of how generative AI tools are being used in the world today:

• Meeting assistance - notetaking, summarization, translation, live captioning
• Text generation​ - drafts of memos, slides, email, etc.
• Audio, video, and image generation 
• Refining text - proofreading, adjusting tone, summarization
• Chatbots - vendor-specific support
• Search engines - AI summary of results

Considerations when using generative AI

Generative AI can be a powerful tool when used responsibly, but there are many important considerations for its use. AI features may be always-on, opt-in, or subscription-based, and opting in to those features may grant broad access to your data. All use of these tools must adhere to ¾«¶«AV policies, including the Acceptable Use Policy and the Data Classification and Handling Policy. Restricted data should not be entered into any system, including generative AI tools, that is not approved for that classification to help avoid unintentional data exposure, regardless of the platform. Use caution what data and documents you share with AI tools; some data should never be shared. 

¾«¶«AV’s contractual agreements with Google, Microsoft, and Zoom protect college data and intellectual property. AI tools provided by these vendors (Google Gemini, Microsoft Copilot, Zoom AI Companion) are authorized for use at ¾«¶«AV. More information about using these tools is available in the “¾«¶«AV Authorized AI Tools with Data Confidentiality Protection” section of these guidelines, below.

Do not upload information to an unauthorized AI tool unless the information is clearly innocuous; even some data that have low sensitivity in the ¾«¶«AV Data Classification policy should not be shared with unauthorized AI tools (e.g., do not upload a list of ¾«¶«AV names and addresses).

We strongly recommend familiarizing yourself with the ethical, privacy, and equity concerns for generative AI, which are covered in detail on the informational page from the Center for Teaching and Learning (CTL).

This document provides guidelines for using generative AI. Given the fast evolution of AI capabilities, it may not address all situations. If you have questions beyond the guidelines given here, please contact the data steward (or the relevant office) who oversees the data in question or the IT help desk (cus@reed.edu). Employees should err on the side of caution when using generative AI tools.

Associated Documents

• Data Classification and Handling Policy - This policy tells you how to handle data in an ethical, professional, and compliant manner.
• Acceptable Use Policy - This policy covers how to use ¾«¶«AV’s technology resources with integrity, while protecting confidentiality and ensuring compliance with applicable laws and regulations.
• Learn more about generative AI from the Center for Teaching and Learning (CTL) 

¾«¶«AV Authorized AI Tools with Data Confidentiality Protection

¾«¶«AV has enterprise-level agreements for a select group of AI tools to ensure the confidentiality and privacy of institutional data, student information, and intellectual property. When you are logged in with your ¾«¶«AV credentials, any data, queries, or inputs you enter will not be used by the vendor to train their underlying AI models. All entered data will not be shared with third parties, publicly disclosed, or otherwise utilized outside of the immediate service delivery context.

• Google - ,
  • Gemini is the recommended chatbot as it best meets our contractual requirements and allows us to provide reliable management and support.
  • Make sure you are signed into your ¾«¶«AV account. Gemini should say “¾«¶«AV” in the top right corner.
• Microsoft -
  • Make sure you are signed into your Microsoft account that is associated with your ¾«¶«AV email. If you have installed Office on a personal computer, you already have an account. If you need a Microsoft account, visit ¾«¶«AV’s Microsoft web page to learn more.
  • Helpful hint: Click your account name in the bottom left corner to verify you are signed in as your ¾«¶«AV account!
• Zoom -
  • Many settings are available but turned off by default.
  • We recommend that you do NOT send meeting summaries to all invitees. You should review meeting summaries for accuracy before sharing them.

Note: Protections provided by these agreements only apply when accessing the tool through your ¾«¶«AV login.

Faculty and staff seeking to acquire or enable new ​A​I products should submit a tech adoption request and IT can assist you in evaluating the privacy, security, and accessibility implications.