How do you train staff in best practices when entering data into an AI tool whether it is ChatGPT, CoPilot, etc.?
Each solution has its own nuances. If you have an account with ChatGPT/Open AI, there is a temporary chat option which will automatically delete your conversation after 30 days and opts you out of your data being used for training. Creating an account with ChatGPT/Open AI also gives you access to manage your privacy. Co-pilot uses encryption and data anonymization techniques to protect user data. The data is not used to train the learning models, and data is encrypted at rest and in transit. Users are also provided with privacy controls.
Whichever tool you decide to use, a cautious approach is best.
Privacy is under the cybersecurity umbrella; therefore, to ensure privacy the first step is to review your cybersecurity practices. Please review the Top 10 Ways to start your Cyber/Privacy plan.
Below is a best practices checklist when implementing AI tools.
- Are your cybersecurity practices in place?
- Strong passwords, MFA, firewall, etc.
- Review permissions and access controls of sensitive data, etc.
- Know what data you have and where it lives.
- Do you have data loss prevention tools in place?
- Create a usage policy gaining insight from all faculty and staff.
- Provide training on expectations of use.
- Do not enter any Personally Identifiable Information (PII).
- Train staff on what is considered PII. Any piece(s) of information that can be assembled to identify is PII.
- Consider the role each staff member has and provide guidance specific to their duties.
- Have practice sessions on entering prompts to obtain the correct output.
- Provide examples of usage within your policy – creating student guides, rubrics, lesson plans, general policies and bulletins, etc.
Following these guidelines will assist users in being successful in ensuring privacy and security when using AI.