This is not a policy. This is a work in progress as produced by the Committee on GenerativeAI adoption at Northfield Mount Hermon (the School).
Privacy and confidentiality
Entering information into AI tools (like ChatGPT) should be considered the same as posting that information on a public website. Expect any information given to an Generative artificial intelligence (GenAI) service will be provided as an answer to another user’s prompt.
This data sharing can happen via:
- information entered via chat prompts
- uploading material (images and files)
- background information provided to improve personalization
Do not share information that is:
- personally identifiable
- Confidential, sensitive, or restricted
- proprietary intellectual property
Do not upload any school data that isn’t publicly available. This includes:
- contracts with budgetary or contact information
- identifiable events with sufficiently identifiable information even if not names (e.g. conduct or vehicle accident reports)
Accuracy and Bias
GenAI services excel at applying advanced statistics to existing data to create new content. However, since GenAI does not understand what it produces, results can be misleading, outdated, or false.
- Do not assume that a GenAI bot is providing accurate information. A human should always validate results.
- Any output from GenAI services should align with the School’s ethical values.
Transparency and disclosure
Until the legal landscape is clearer and there is a formal policy, we recommend disclosure whenever GenAI is used to produce published work.
- When a work includes a full sentence or more verbatim from an AI-assisted technology, we recommend disclosure similar to:
- “Disclosure: This [piece, article, text] was edited with assistance from AI technology.”
- When a production includes significantly altered images, audio, or video via GenAI tools, we recommend disclosure similar to:
- Disclosure: (Original credit when applicable). A portion of this [image, audio, or video] has been enhanced using AI technology.”
These disclosures can be in the footnote of a document or text, e.g. admissions pamphlet or handbook.
Legal
The School’s AI stance falls within the scope of the Written Information Security Policy (WISP) in the Employee Handbook. The WISP is designed to ensure the security, integrity, and confidentiality of non-public information. NMH is required to protect its information against anticipated threats and guard against unauthorized access or use.
- Use of GenAI services must comply with law.
- Copyrighted materials, or any portion, should not be uploaded to an AI service, even if you or the School owns the content.
- In most cases, content created by generative AI cannot be copyrighted.
Tools
Without a legal contract stating otherwise, any information provided to public generative AI tools is considered public and discoverable. Currently, NMH does not have any specific contract or agreements with generative AI tools or services.
This document borrows significantly from guidelines and policies of other institutions, some of which may be behind institutional logins. In particular, the authors note contributions from Choate Rosemary Hall, Phillips Exeter Academy, University of Notre Dame, University of Rochester, Yale University, CommonSense.org, Eric Hudson, and Lance Eaton.