Guardrails and Privacy
Set clear team rules for AI use, with extra care for the data of young people.
Objective
By the end of this lesson, you can write a one-page AI use policy for a team and teach it with real examples.
Why teams need written rules
When AI use is not written down, every person makes up their own rules. One person pastes client notes into a chatbot. Another refuses to use AI at all. Nobody knows what is allowed.
Guardrails are the written rules that keep tool use safe. They protect the people you serve and the team itself.
A national example: a lawyer and fake cases
In 2023, lawyers in a federal case in the United States, Mata v. Avianca, filed a brief that cited court cases ChatGPT had made up. The cases did not exist. The judge sanctioned the lawyers. The tool did not check its own work, and neither did the people.
Teach this as a rule: AI output is a draft until a person checks it.
A one-page AI use policy
Keep it short enough to fit on one page. Here is a starter.
| Rule | What it means |
|---|---|
| Check everything | Every fact, name, date, and number is verified before it is used |
| No private data | No names, contact details, case notes, or health or money details go into AI tools |
| Extra care for minors | No information about anyone under 18 goes into AI tools. Ever. |
| Say when AI helped | If AI drafted something important, the reviewer knows |
| Approved tools only | The team uses tools the organization has reviewed |
| A person decides | AI never makes decisions about who gets services, jobs, or spots |
The federal NIST AI Risk Management Framework is a free, public guide for organizations that want to go deeper. You do not need it to start. One page of clear rules is enough for a small team.
How to teach the policy
Do not just send it out. Run it through real cases:
- Read each rule out loud.
- Give the team three short scenarios. For each, ask: allowed, allowed with care, or not allowed?
- Update the policy when a scenario shows a gap.
Sample scenarios:
- A staff member wants AI to summarize public grant guidelines. Allowed. Check the summary against the source.
- A volunteer wants AI to rewrite a parent's email that includes a child's diagnosis. Not allowed. It contains private health details about a minor.
- A team lead wants AI to draft a flyer for a free workshop. Allowed with care. Check every date and detail.
Reflection
Go deeper: AI training that sticks: teach the real task, not the tool
Discussion
Ask. Answer. Build together.
Loading…