Secure generative AI for the Harvard community
The AI Sandbox provides a secure environment to explore Generative AI, minimizing security and privacy risks. Data entered into the Sandbox will not be used to train any vendor's large language models (LLMs).
The platform offers a single interface with access to the latest LLMs from OpenAI, Anthropic, Google, and Meta.
The AI Sandbox includes several tools to enhance your AI projects:
- Data visualization – Generate charts and graphs to visualize data insights.
- Multi-file upload – Upload multiple files to streamline data processing.
- Image generation – Create visual content based on text prompts.
Use these links to request access and use the AI sandbox:
- Request access to the AI Sandbox.
- Access the Harvard AI Sandbox if you already have access.
-
Eligibility
Available to Quad-based faculty and staff, and students as part of coursesFaculty StaffThe Harvard AI Sandbox is available to:
- HMS and HSDM Quad-based faculty and staff – Faculty and staff at the Harvard Medical School (HMS) and Harvard School of Dental Medicine (HSDM) Quad can request access.
- Students as part of a course – Course instructors can also request access for students in their Canvas course.
-
Security
Approved for up to Level 3 dataLevel 3 PrivateThe Harvard AI Sandbox is approved for up to Medium Risk Confidential (Level 3) data. Data you enter will not be used to train large language models.
-
Support
HMS provisions users, HUIT provides ongoing supportProvisioning SupportThe HMS IT Teaching and Learning Technologies team provisions user accounts for Harvard Medical School (HMS) users in the Harvard AI Sandbox. To request access, use the AI Sandbox access request form.
Harvard University Information Technology (HUIT) is responsible for maintaining and supporting the Harvard AI Sandbox. For support with issues other than access requests, contact HUIT for help.