K-12Issue 333 records
K–12
Tonight's K-12 signal is the fine print behind student-AI restrictions. Frederick County's adopted Policy 124 text confines any student AI use to FCPS-approved tools reachable only on the district network and monitored, bars staff from using AI to evaluate student work, and preserves IEP/504 access. Utah's draft model policy, posted for the October board, becomes the default for any LEA that does not adopt its own, prohibits generative AI from independently grading or making high-stakes determinations, and sets consent rules for high-school AI sandbox courses. A new preprint warns the approved-tool model rests on weak vendor paperwork: a third of 48 audited EdTech platforms give no meaningful AI disclosure.
- Evidence records
- 3
- Cross-source patterns
- 2
- Evidence classes
- 2 standards or public-body guidance1 academic research
- Outcomes
- 2 cautionary1 emerging
- Source freshness
- 2 new this fortnight1 recent
- Research completed
- 2026-10-10
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Approved-tool lists and human-only grading are becoming the operative K-12 AI controls
Frederick County's Policy 124 and Utah's draft model policy both restrict use to district/LEA-approved tools and bar AI from evaluating or independently grading student work, while preserving disability-related access.
Operating questionDo we have an approved-tool inventory and a technical control ensuring AI cannot grade students without human decision?
Supporting evidenceFrederick County Public Schools Board of EducationUtah State Board of Education
Approval gates depend on AI disclosures vendors often do not provide
Policies now hinge on approving tools, but the EdTech privacy audit finds a third of platforms lack meaningful AI disclosure, so approval needs contract and technical checks beyond vendor policies.
Operating questionWhat AI-specific disclosure and breach terms must a vendor supply before a tool enters the approved list?
Supporting evidenceMeghna Manoj Nair and Rachel Greenstadt (arXiv preprint authors)Frederick County Public Schools Board of EducationUtah State Board of Education
Evidence ledger
Full record · every source keeps its link and limitations
Frederick County Policy 124 text: PreK-12 student AI prohibition limited to approved, network-bound, monitored tools
The adopted policy text goes beyond the headline ban already archived via news coverage: it prohibits student AI use PreK-12 except for AI-literacy instruction, course/extracurricular requirements, or superintendent approval; confines any use to FCPS-approved tools reachable only on the secure FCPS network and monitored by FCPS; bars staff from using AI to replace direct instruction or evaluate student work; mandates AI access where an IEP or 504 plan requires it; and directs regulations across ten areas including human oversight, data security and procurement, plus an advisory group. Why now: the primary text (adopted Oct 7) was not previously archived and specifies the operating constraints.
Why it matters, evidence and limitations
- Why it matters
- Gives districts and vendors the exact clauses a large Maryland district will enforce, rather than press summaries.
- Evidence and measured results
- Primary policy document (5 pages) read in full; no outcome data — it is a governance rule, not an evaluation.
- Limitations and uncertainty
- Policy text only; regulations, approved-tool list and enforcement are pending. Board elections on Nov 3 could alter direction (reported elsewhere, not in this document).
Utah draft model policy: default policy for non-adopting LEAs, no AI grading, parental-consent AI sandbox courses
USBE's draft Instructional Technology and AI Model Policy and Implementation Resource (Draft 2, October Board) implements 53G-7-228/-229, 53G-7-1401 to -1403 and Rule R277-334. It states that if an LEA fails to adopt its own policies the model policy becomes its default; requires LEAs to prohibit generative AI from independently grading student work or making high-stakes determinations, train staff and monitor compliance; requires written parent notification for instructional generative AI; and sets conditions for optional grades 9-12 AI Sandbox courses (CTE only, parent notification, written consent, voluntary). It adds a students-with-disabilities section on AI and the IEP process and appendices including an approved-tools transparency template. Why now: this companion model policy is newly posted for the October board; the archive holds only the R277-334 rule draft.
Why it matters, evidence and limitations
- Why it matters
- Shows the concrete local-policy template Utah LEAs will inherit by default, complementing the archived rule.
- Evidence and measured results
- Draft document (~3,100 text lines) inspected via table of contents and the AI, sandbox, LEA-responsibility and model-policy sections; draft redlines visible.
- Limitations and uncertainty
- Draft with tracked bracketed edits; final board action not observed. No outcome evidence.
EdTech privacy deferral study: one-third of 48 platforms give no meaningful AI disclosure despite visible AI features
Mixed-methods preprint: 12 interviews with EdTech professionals plus a privacy-policy audit of 48 platforms on five dimensions (mean Cohen's kappa 0.781). Privacy is recognized but deferred; responsibility is delegated to cloud providers, documents or downstream institutions. 33% of platforms make no meaningful AI disclosure despite visible AI features; 73% offer only generic accountability and breach-response language. K-12 platforms do better on children's consent where regulation is explicit, but not on AI governance or accountability.
Why it matters, evidence and limitations
- Why it matters
- Directly bears on district approved-tool processes now being mandated (Frederick County, Utah): vendor policies may not disclose AI use well enough to approve tools.
- Evidence and measured results
- Abstract page inspected; full paper not reviewed; preprint not peer reviewed.
- Limitations and uncertainty
- Preprint, not peer reviewed; only the abstract was inspected; 12-interview sample; platform selection and K-12 share not verified.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Standards or public-body guidance
- Normative or advisory guidance from a standards body or public institution.
- Academic research
- Research produced through an academic institution or peer-reviewed venue.