Data Security
The different models used by generative AI platforms draw from large datasets to respond to prompts and provide content (e.g., images, text, code, etc.). Information entered into public or personal AI services may be used to train its model, depending on provider’s policies and your consent settings, and this increases the risk of unauthorized exposure or use in ways the University cannot control.
The University establishes contracts with approved providers and tools to help comply with institutional policies, compliance regulations, and state/federal laws. Under those agreements, your prompts and files are not used to train vendor models, content is retained according to the agreement and university record retention rules, and administrative access is limited to what university policy and legal obligations require.
You play a role in protecting your information and institutional data.
- Use UConn-sanctioned tools for university work.
- Make sure you are logged in to your official UConn licensed account.
Know what kind of data you’re using.
According to the Data Classification Policy, the three classifications of data – public, protected, and confidential – must all be safeguarded at some level.
- Protected or confidential data may only be stored or processed within university-contracted AI platforms.
- Specific legal or contractual controls may completely prohibit certain confidential data from being used even within licensed AI tools; users should verify with data owners.
Additional Guidelines and Considerations
Academic Use
Considerations and resources for instructors from the Center for Excellence in Teaching and Learning: https://cetl.uconn.edu/generative-ai/
Use of AI in communications at UConn
Outline of acceptable uses of AI to develop visual and written content and practical guardrails from University Communications: https://brand.uconn.edu/resources-tools/ai-guidelines/
Accessible Content
Digital content created with AI must comply with the University’s accessibility policy and the requirements of the Title II of the Americans with Disabilities Act. Using a supported platform does not guarantee accessibility; content should be reviewed for compliance.
Developer & Administrators
Review the security advisories and publications from the Cybersecurity and Infrastructure Security Agency (CISA) and international partners (such as the UK NCSC) for AI developers and systems administrators building agents, deploying AI models, and securing data pipelines.