Use AI Carefully

Some uses require stronger safeguards, clearer review roles, and more restraint.

These uses are not automatically off-limits, but they should not be treated casually.

What this page helps staff do

Recognize when an AI use is moving closer to evaluation, recommendation, or protected information and needs a slower, more deliberate process.

What still requires human review

Whether the task should use AI at all, whether the tool is appropriate, and whether the output could shape a student's opportunities unfairly.

IEP-related language

Drafting may be possible, but final language must stay under human control and team review.

Student data summaries

This includes present levels, qualitative work evaluations, transition assessment narratives, and other summaries based on real student records. Even when names are removed, these summaries may still contain indirect identifiers or details that could make a student recognizable, so they belong in a slower, more careful category.

Assessment summaries

AI should not replace interpretation of transition assessments or contextual judgment.

Placement recommendations

AI should not decide services, placements, or readiness.

Student-facing chatbots

Direct student interaction raises new concerns about privacy, reliability, and over-reliance.

Stored prompt tools

Tools that retain or train on prompts require closer review of privacy and policy alignment.

Stop and check

If the task involves identifiable student information, recommendations about a student, or anything that could shape services or placements, pause and verify the safeguards, approval path, and review responsibility first.

  • Do not paste protected records into a tool by default.
  • Do not assume “de-identified” only means removing the student's name.
  • Do not paste narrative details that could still point back to a specific student, school, class, family, or work site.
  • Do not use AI to quietly narrow placement or service options.
  • Do not treat generated recommendations as if they came from the team.

Safer version

Use AI to organize or summarize genuinely de-identified, minimal notes in an approved tool for drafting support only.

Use carefully version

If the summary is based on a real student's record and could still point back to that student, treat it as a careful-use task, not a routine-safe use.

Concrete example

Initials are not enough on their own

Even in a district-approved tool, replacing a student's name with initials does not automatically make a prompt safe. Details such as a disability, a particular job site, a teacher, a school, or an unusual set of needs can still make the student recognizable to people who know the context.

Still too identifying

“J.D. is an 18-year-old at Lincoln High with an intellectual disability. He needs extra processing time while learning tasks at the Green Street Cafe. Summarize his work evaluation.”

Safer starting point

De-identified notes: “Completes familiar multi-step tasks independently after a visual model. Asks for clarification when a task changes. Benefits from a brief written checklist when learning a new routine.”

Prompt: “Organize these notes into strengths, support needs, and next instructional questions for a draft work-evaluation summary. Do not make recommendations or conclusions.”

Use only the minimum information needed, follow district approval and policy, and review the draft against the original record before using it. An approved tool may allow specific student information under local rules, but approval is not blanket permission to paste every detail from a record.