K-12Issue 475 records
K–12
New York City's student AI restrictions are running into a definition problem. Chalkbeat obtained the 38 ed tech products the city flagged, including reading and math curriculums and the department's own Microsoft-built chatbot; several vendors say they use no prohibited AI, and the department has not explained its criteria. A privacy advocate's review finds four of the city's five AI high school pilot vendors lack public terms barring product-improvement use of student data. Elsewhere, districts are writing their own controls: Sarasota County proposes parental opt-in, closed systems and a ban on companion-style AI, while Cherry Hill adopted five assignment-level AI tiers with one monitored platform and no opt-out. In Maryland, Anne Arundel, Baltimore City, Baltimore County and Howard missed the state's policy deadline. No source reports measured learning outcomes.
- Evidence records
- 5
- Cross-source patterns
- 2
- Evidence classes
- 5 independent reporting
- Outcomes
- 2 cautionary2 emerging1 mixed
- Source freshness
- 5 new this fortnight
- Research completed
- 2026-10-24
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Districts are deciding AI eligibility on vendor descriptions they cannot easily verify
NYC flagged 38 products and vendors dispute whether they use student-facing generative AI, with no public criteria; an advocacy review of NYC's pilot vendors finds public terms that allow product-improvement use and argues the department cannot verify claims without code access. Sarasota's proposal answers with contract terms and closed systems, but enforcement still rests on vendor attestations.
Operating questionFor each student-facing tool, what evidence beyond the vendor's description shows which AI features exist, where student data goes and whether it trains models?
Supporting evidenceNew York City Public Schools (reported by Chalkbeat New York)Parent Coalition for Student Privacy / Class Size Matters (for the AI Moratorium Coalition)Sarasota County Schools (reported by The Sarasota Record)
Districts are splitting on parent consent for student AI
Sarasota proposes parental opt-in with a comparable non-AI alternative; Cherry Hill dropped a parent opt-out over equity concerns and relies on platform monitoring; Anne Arundel parents and teachers are asking for a ban while the policy is drafted. Consent design is becoming a central policy choice, with no evidence yet on which approach works better.
Operating questionWhich consent model (opt-in, opt-out, none with monitoring) fits the district, and how will it be enforced in rostering and offered alternatives?
Supporting evidenceSarasota County Schools (reported by The Sarasota Record)Cherry Hill Public Schools (reported by Cherry Hill Courier)Anne Arundel County Public Schools (reported by The Baltimore Banner)
Evidence ledger
Full record · every source keeps its link and limitations
Chalkbeat obtains NYC's list of 38 ed tech products flagged under its student AI restrictions; several vendors say they don't use the prohibited AI
After more than a month of requests, Chalkbeat obtained the Education Department's list of 38 previously allowed products whose AI components the city said it would discontinue or disable under its new student-facing AI restrictions. The list includes widely used reading and math curriculums, student assessment tools and the department's own chatbot teaching assistant built with Microsoft and used in six schools last year. Amira, an AI reading tutor used in more than 200 city schools, was banned outright because officials said it cannot function without AI. Several companies, including HMH (Read180) and the translation tool Wordly, said their products do not use student-facing generative AI and did not know why they were flagged. The department said restrictions apply to student-facing generative AI features, not necessarily whole products, and that it is working case by case on whether features can be disabled, but did not explain how it decided which tools met the definition. Some vendors disabled features quickly: All In Learning turned off AI-assisted grading, and Magma Math disabled AI hints on September 4 and was cleared on October 5. Coursemojo and Gemini are also banned but not on the list.
Why it matters, evidence and limitations
- Why it matters
- The largest U.S. district's restriction is being implemented product by product, and the list shows how hard it is to define and verify 'student-facing generative AI' across a large ed tech portfolio, a problem every district with an AI policy will face.
- Evidence and measured results
- Investigative reporting based on an obtained district list and company responses; no district documentation of the vetting method was released.
- Limitations and uncertainty
- Reporting covers the list and company statements; the list is not exhaustive per officials; the department's criteria and testing process are undisclosed. Earlier NYC moratorium coverage is archived; this source adds the obtained product list and vendor disputes.
Class Size Matters / Parent Coalition for Student Privacy: four of five NYC AI high school pilot vendors' public policies allow product-improvement use of student data
Leonie Haimson's analysis reviews the public privacy policies and posted privacy addenda of the five announced NYC AI high school pilot vendors. It says Playlab, Quill and Edia state that personal student data may be used to improve their products or services, which the author argues is a commercial use prohibited under New York's Education Law 2-d regulations. Intel AI-Ready Schools has no public-facing privacy policy and no posted addendum; Intel's general privacy notice allows use for advertising and marketing. Brisk explicitly says it does not use student or teacher data to train or fine-tune models, but the analysis says its Brisk Boost student chatbot appears to contradict the department's statement that companion chatbots are prohibited. For the posted addenda, the analysis says none explicitly bars product-improvement use, requires data minimization or sets a deletion date, and Playlab's listed contract term ran only from February to June 2024. It also says the department cannot verify vendor claims without access to code or running the tools on its own servers.
Why it matters, evidence and limitations
- Why it matters
- A line-by-line privacy review of the actual vendors in the largest district's AI pilot, and newly relevant tonight because Chalkbeat's October 23 list shows the department's AI vetting is contested and quotes the same author.
- Evidence and measured results
- Advocacy analysis of public documents; legal interpretations are the author's; no response from vendors or the department is included.
- Limitations and uncertainty
- Published October 13, before the last run, but not archived; selected because of today's NYC list. Advocacy source with a stated position; legal conclusions are not adjudicated; vendor contracts may contain terms not public.
Sarasota County Schools (FL) advertises AI policy revisions: parental opt-in, no AI decisions without human review, closed systems, ban on companion-style AI
The Sarasota County School Board approved revisions to Policy 4.24 (AI Acceptable Use) and Policy 3.43 (Internet Safety) for advertising on October 20; notices posted October 21 give 21 days to request a hearing, and final approval is pending. Students could not use AI unless a parent opts them in and the teacher allows it, must cite AI content, and must be allowed to show mastery without AI. AI could not make decisions without human review, including grading, discipline, threat assessments, counseling referrals and ESE referrals. By July 1, 2027, teachers and administrators must complete district AI training before using a new AI tool with students. Approved tools could not sell, monetize, profile or exploit student data to train commercial models; U.S. data storage gets priority; undisclosed behavioral monitoring, social scoring and psychological profiling are banned. Supplemental platforms must run on closed systems, align tutoring to Florida's B.E.S.T. standards and keep interaction records at least 30 days for parent access. The internet-safety policy bans AI designed to meet students' social or emotional needs or simulate friendship, and requires a comparable non-AI alternative for students not opted in.
Why it matters, evidence and limitations
- Why it matters
- A proposed district policy that combines parental opt-in, human review of consequential decisions, vendor data terms and a ban on companion-style AI, giving vendors concrete requirements in a large Florida district.
- Evidence and measured results
- Local news summary of proposed policy text and public notices; policy not yet adopted; no usage data.
- Limitations and uncertainty
- Proposed, not adopted; based on a news summary of the policies and notices, not the policy text itself; final wording may change after hearing requests.
Cherry Hill (NJ) approves five-level AI use policy: default Level 0, SchoolAI as sole approved platform, parent opt-out dropped over equity concerns
Cherry Hill's school board approved a policy on October 13 for the district's 10,919 students that assigns every assignment an AI use level from 0 to 4; teachers must state the level in writing, and the default is Level 0. Level 4 allows AI throughout an assignment if students document prompts and show their thinking. Pre-K and kindergarten students may not use AI independently, second graders may do whole-class activities under supervision, grades 4-5 may use AI on specific assignments but not to generate original work, and high schoolers can use all levels at a teacher's direction. High-stakes tests default to no AI. SchoolAI is the approved platform; the superintendent said its closed system lets teachers monitor interactions in real time. The board dropped a parent opt-out after equity concerns, with officials saying monitoring addresses safety; a parent called the policy half-baked. An AI coordinator maintains the approved list and leads an annual review by April 1.
Why it matters, evidence and limitations
- Why it matters
- An adopted, grade-banded policy that relies on one monitored platform instead of parent opt-out, a contrast with opt-in models elsewhere and with neighboring Haddonfield's elementary ban.
- Evidence and measured results
- Local news report citing The Philadelphia Inquirer and district statements; no usage or outcome data.
- Limitations and uncertainty
- Secondary report of an October 13 approval first reported by the Inquirer (not opened); earlier Cherry Hill community-session coverage is archived, but this is the adopted policy. Superintendent's safety claim about monitoring is untested.
Baltimore Banner: Anne Arundel among Baltimore-area districts missing Maryland's Oct. 22 AI policy deadline; parents and teachers urge a ban
The Banner reports that Anne Arundel County Public Schools is among a handful of Baltimore-area districts that missed Thursday's state deadline to adopt an AI policy. Its draft requires AI tools to be vetted for educational value before adoption, says AI may not replace employees' duties, and directs the district to set expectations and limits for student use. Parents and teachers at this week's board meeting, including a Schools Beyond Screens coalition parent, a middle school art teacher and a school librarian, urged a ban until effects are studied; no supporter spoke. A board member questioned the legislative deadline given the complexity. Baltimore City, Baltimore County and Howard County have not yet introduced AI policies, though Howard gave high school students Google Gemini earlier this year. The superintendent is holding community meetings over the next few weeks, and the board plans to adopt a policy by mid-December.
Why it matters, evidence and limitations
- Why it matters
- Shows the state mandate's first deadline passing with several large Maryland systems out of compliance, including one that already gives students Gemini, and organized parent pressure for bans.
- Evidence and measured results
- Regional news report from a board meeting; quotes from speakers; no compliance count from the state.
- Limitations and uncertainty
- Anne Arundel's first reading is already archived from October 21-22 coverage; this new source adds the missed-deadline status of Baltimore City, Baltimore County and Howard County and the board's December target. No state data on compliance; speakers' views are not representative.
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.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.