SPH Generative AI Sandbox Now Available

Dear SPH Staff,

With interest in Generative AI very high, we are pleased to announce the launch of a SPH Generative AI sandbox for secure experimentation.  This sandbox may be used to explore large language models (LLMs), search for content, create/edit/summarize content, assist with code writing/debugging, analyze information and much more.

SPH AI Sandbox

SPH’s sandbox may be found here, and is part of Harvard University’s larger AI sandbox.  Additional information on getting started with the Sandbox may be found in the email below from Harvard University Information Technology.

SPH Survey

To assist SPH and the University in planning for future Generative AI needs, we ask that you complete a brief survey, Staff GAI survey, by December 15.


Keep an eye out for SPH workshops and informational sessions on experimenting with generative AI in teaching and learning, administration and operations, and research.We anticipate that the sandbox will be available through the end of the academic year.

Kate Calvin, Executive Dean for Administration
Erin Driver-Linn, Dean for Education
Deane Eastwood, Chief Information Officer

Dear SPH Staff,

We’re pleased to announce that the pilot version of the Generative AI Sandbox is now ready for you to use.

As a reminder, the AI Sandbox has been developed by HUIT in collaboration with VPAL, the FAS Division of Science, and colleagues across the University, to enable Harvard community members to securely access Large Language Models (LLM). The AI Sandbox offers a single interface that enables access to four different Large Language Models (LLM): Azure OpenAI GPT-3.5 and GPT-4, Anthropic Claude 2, and Google PaLM 2 Bison. It provides a “walled-off,” secure environment in which to experiment with generative AI, mitigating many security and privacy risks and ensuring the data entered will not be used to train any public AI tools.

During this first pilot period, the AI Sandbox:

  • Is approved for use with Medium Risk Confidential data (L3) and below.
  • Is free to use.
  • Features a simple log in process using your existing HarvardKey credentials.
  • Cannot be custom trained on a corpus of materials.
  • Will experience occasional errors and interruptions in service, which may reset your session.

Currently, API access is not available through the AI Sandbox. Please visit HUIT’s website on exploring Generative AI tools for more information about OpenAI API usage.

About Generative AI

Harvard has launched an informational page about Generative AI at https://huit.harvard.edu/ai/about.

Before you use the AI Sandbox:

Please review “Getting started with the AI Sandbox,” which includes guidelines, terms of use, tool instructions, and ideas for prompts.

To access the AI Sandbox:

  1. Go to https://hsph.sandbox.ai.huit.harvard.edu/
  2. When prompted, log in with your HarvardKey username and password.


Your feedback will help us develop the AI Sandbox further and inform future University investments in AI technology. Please use this survey to submit your comments.


The pilot version of the AI Sandbox contains elements that are not fully accessible. If you or any participants need accessibility assistance, the following resources are available:

Student Local Disability Coordinators (LDCs): https://accessibility.harvard.edu/student-coordinators.

If you are unable to determine who the appropriate contact may be, University Disability Resources can be contacted via accessibility@harvard.edu or Digital Accessibility Services via digitalaccessibility@harvard.edu.

Thank you,
Academic Technology
Harvard University Information Technology