A Guide to the Harvard AI Sandbox
What is the Harvard AI Sandbox?
The Harvard AI Sandbox provides a secure environment for you to experiment with generative AI. Please note, you should use the sandbox instead of other accounts. The sandbox is here for you to responsibly experiment with Generative AI tools.
How can I access the Harvard AI Sandbox?
You can access the Harvard AI Sandbox by accessing the sandbox site and signing in with your HarvardKey. You can also request the AI Sandbox for the students in your course by requesting access via HUIT.
Tips for Getting Started with Teaching with AI
Tip #1: Share AI policies for use & citations
State your AI policy in your syllabus and tell students they must disclose the use of AI in their work. Harvard Library has guidelines for how to cite AI for research and scholarship.
Tip #2: Use a variety of tools
Practice using 7 different AI tools using the Harvard AI Sandbox. Here is a comparison of Gen AI tools including text-based tools, AI assistants, image-based tools, and developer tools.
Tip #3: Check for accuracy
Assume that any information generated by AI may be fabricated (or “hallucinated”). Validate the content from a legitimate source.
Be mindful of bias. AI uses data created by humans, which can have inherent bias.
Tip #4: Ensure privacy & confidentiality
If you're using GenAI for class purposes, you should do so in the Harvard-approved AI Sandbox, which will ensure privacy up to level 3 data. The list of Gen AI tools from HUIT includes the level of confidential data for approved usage in the 4th column of the table.
Confidential information can not be loaded into Gen AI systems, since there is no expectation of privacy or confidentiality. Faculty must get documented permission from students before putting original student content into any generative AI tool, and students should be made aware of the risks of entering their original work into such tools.
Source: Harvard College Office of Undergraduate Education AI Guidance & FAQs