Lifetime Citizen Portal Access — AI Briefings, Alerts & Unlimited Follows
Mississippi district says local AI server eased grading and protected student data
Loading...
Summary
The superintendent of the Pearl Public School District described using a district‑hosted AI instance to automate parts of grading, protect student privacy, and provide individualized student plans; he and witnesses said in‑district deployments can reduce FERPA risks but raise cost questions for smaller districts.
Chris Chisholm, superintendent of the Pearl Public School District in Pearl, Mississippi, told the subcommittee his district built and runs its own AI server to avoid sending student‑identifiable data to third‑party models and to speed teacher workflows.
Chisholm said an English teacher who had planned to retire because grading took three to four hours nightly returned to the classroom after the district built an "agent" that uses the state writing rubric to pre‑scaffold grading. "We created an agent for her…now she doesn't have to take those papers home and grade them," he said, adding that the AI is a tool the teacher controls rather than an automated replacement.
Chisholm also described using student assessment data to generate individualized short, medium and long‑term plans: "We upload a couple of PDFs, we tell it what we want it to do…it gives that kid a short term, medium term and long term plans that are developed around that individual child." He said the district's private server eliminates internet‑facing FERPA concerns.
On privacy, Chisholm said districts using public models should teach staff to "scrub" identifiable information before uploading it to external services: "If you're using a public entity…you can't put student identifiable information on it…teaching people to scrub files, that's the biggest thing." He acknowledged the server approach is costly but said it can be more cost‑effective than paying for tokens or subscriptions to commercial models at scale.
Superintendent Chisholm said Pearl also engages regularly with major AI vendors and has scripted models to prefer district‑identified sources. He recommended careful vetting of vendors, clear training for teachers on data handling and shared governance over how AI tools are used.
No federal or state mandates were announced; Chisholm’s testimony illustrated one district’s local approach to the tradeoffs among privacy, cost and control.
