Dear Master’s Students, We have all witnessed the rapid advancements in artificial intelligence technologies over the past several years, including rollout of large language models (LLMs) like ChatGPT. We know that these can be useful tools in healthcare and education, when used appropriately. We are therefore happy to announce that Harvard Medical School will provide all HMS master’s students access to the HUIT AI Sandbox effective immediately for the 2025-2026 academic year. You can access the AI Sandbox here by clicking “sign in” and logging in with your HarvardKey credentials. Before using the AI Sandbox for course work or academic writing, please confirm the specific AI use policies governing the academic work in your program and individual courses with your program directors and/or individual course directors. If approved for use, remember that you must disclose use of AI tools and cite them appropriately. As detailed further below, while you may upload course materials to the AI Sandbox (unless explicitly disallowed by your course director), you should not upload materials to “free” genAI tools like Google Gemini of free ChatGPT. Finally, we encourage you to engage your peers, instructors, and program directors in conversations about how we as a community can productively engage with generative AI in our academic and professional activities without short-cutting essential learning processes. Additional details about this resource are given below. We are incredibly excited to offer this Sandbox to all HMS students. While we work to get formal evaluation and feedback systems in place, if you have any comments or concerns, please feel free to reach out to Taralyn_Tan@hms.harvard.edu.
Additional Information:
The AI Sandbox contains multiple AI models, including the latest versions of ChatGPT, Claude, Llama, Gemini, and DALL-E, as well as smaller models. It also has the ability to directly upload multiple documents, to generate charts and graphs to visualize data insights, and create visual content based on text prompts. Our goal is to encourage both educational experimentation and the development of a community of learning. With that in mind, we are providing the following policies to guide use of the HUIT AI Sandbox: 1. Privacy. The AI Sandbox has many privacy protections, but it is not HIPAA-
compliant. No identifiable patient information should ever be entered into the Sandbox. Similarly, patient notes, even if stripped of identifiers, should not be uploaded. We also caution you about uploading any sensitive or embargoed research data. However, the sandbox does have many other privacy protections in place, and personal information about yourself, including items such as course evaluations, as well as curricular materials from Harvard Medical School, are safe to upload. Unlike public models, none of the information that you put into the Sandbox can be accessed outside of the 1