AI Security: What to do, what not to do

Below are general guidelines and safety best practices” for configuring and interacting with AI tools. If you regularly work with sensitive data and use AI, talk to your supervisor and/or team, and consult with LTS if you have any questions.

Best practices (do’s and don’ts)

1. When using AI, do not share or upload protected data such as LINs, social security numbers, FERPA-protected information, etc. For more information see guidelines for Data Classification 1 and Data Classification 2.

2. ChatGPT. Do configure your AI tool to disable “training” - a setting that determines whether your data can be used to train the AI model, and thus becomes part of the publicly available content.

   1. In OpenAI ChatGPT Click Profile (upper right) → Settings → Data Controls → Improve the model for everyone → Off

      Open

      

      Turn off “improve the model for everyone”

   2. Claude. Do monitor Anthropic Claude practices. Currently they state “We will not use your prompts and conversations from Free Claude or Claude Pro to train our models. We use your prompts and conversations to train our models: (1) if you give us explicit permission by submitting feedback through the thumbs up/down feature or by reaching out to us with a request, and (2) where your prompts and conversations are flagged for trust and safety review, we may use or analyze those conversations to improve our ability to detect and enforce Usage Policy violations.

   3. Gemini. In Google Gemini, do log into gemini.google.com with your Lehigh credentials. Lehigh’s agreement with Google offers some protection for data we utilize/share in AI tools.