Issue 09 · 完整试读Schools set AI boundariesWhat can stay? From restrictions to teacher learning and high-touch support, the next adoption test
2026 Issue 09No. 009
Issue 09 · FULL PREVIEW
Schools Are Writing AI Boundaries. What Can Stay?
From restrictions to teacher learning and high-touch support, the next adoption test
School policies are specifying which grades and assignments may use AI, while training organizations and universities are placing AI literacy inside teacher communities and STEM education. As boundaries become concrete, products must serve an allowed, supervised task with a named person responsible for it.
Reporting window
to
Judgment
The more specific the rules, the more a product must turn age, task, teacher responsibility, and exit into an executable classroom workflow.
Evidence base
District policy, teacher courses, a regional symposium, and a university initiative
ISSUE SUMMARY · ISSUE SUMMARY
This week’s signals come from both limits and support. Frederick County published a PreK–12 policy that bars student AI use for substitution and grading; NAAIC is running curriculum deep-dives for educators; LACOE brought human-centered practice to a regional symposium; and MIT launched a longer-term STEM initiative for America. As rules tighten, a product’s reason to enter school becomes smaller, more specific, and easier to test.
COVER STORY · COVER STORY
When boundaries enter policy, tasks become the entry point
Frederick County Public Schools records that its board adopted Policy 124 on October 7. The policy bars K–12 students from using AI to replace their thinking, complete assignments, or grade work. Explicit AI-literacy teaching, curricular activities, and IEP support are exceptions.[1]
The rule divides “allowed” use by grade, task, and responsibility. A product that generates an answer may be unsuitable for one assignment. A product that helps a teacher teach checking, citation, and reflection may fit an allowed classroom step. The policy page gives no score improvement or purchasing result, so it cannot establish market size.
At the same time, NAAIC’s Day of AI curriculum deep-dives schedule eight live hours and a community of practice for 25 educators, while LACOE’s regional symposium brought 130 participants from more than 40 districts.[2][3] Support is growing, and teacher time is part of the adoption cost.
WEEK IN REVIEW
This week: policy boundaries and teacher support become more specific
Frederick County adopted a student AI policy. It spells out bans on substitution, assignment completion, and grading while preserving instructional exceptions for AI literacy, curriculum, and IEP support.[1]
NAAIC scheduled curriculum deep-dives for educators. The cohort has 25 educators, eight live hours, and a community of practice. It is a training arrangement, not a classroom-outcome study.[2]
LACOE published a regional symposium account. It reports more than 40 districts and 130 participants and emphasizes human-centered practice. Attendance is not regional procurement or a student result.[3]
POLICY
Rules narrow product choice to a concrete teaching task
Policy 124 offers a clear entry point: students may learn tools in teacher-directed AI-literacy activities but may not present generated work as their own. For a product, permissions, prompts, citation, and teacher review need to form one workflow. A chat window alone cannot answer what an assignment permits.[1]
District policies, age bands, and IEP responsibilities vary. Providers need configurable rules, audit records, and an exit path that schools control; one district’s exception is not universal permission.
PEERS
High-touch support makes teacher time part of the product
MIT for America targets K–community college and emphasizes STEM, teacher education, and face-to-face support.[4] MIT RAISE’s updates also place AI literacy, teacher learning, and community collaboration together.[5]
These programs’ strength is a person helping a school complete the work; the cost is local coordination and training. A scalable product must show which steps can be reused by software and which need local staff. High-touch service may improve adoption quality while compressing margin.
OPPORTUNITY
Build the smallest loop around an allowed task
Start with one assignment rule. Choose a grade and task, specify what students may do, what teachers inspect, and how outputs are marked. Record the time for one complete workflow.
Put teacher learning inside delivery. Training is not a launch announcement. Examples, co-planning, review, and escalation should help teachers decide independently whether to continue.
Design an exit for policy changes. Districts should be able to export records, turn off features, and keep student work as rules change. Reversible settings lower adoption risk and reduce lock-in disputes.
NEXT WEEK
Watch next: enforcement, refresher learning, and task results
Look for Frederick County implementation guidance and teacher feedback, classroom examples from NAAIC, shared workflows among LACOE districts, and defined support outcomes in MIT’s initiative. If boundaries are clear but task results remain unobservable, the product case needs to narrow further.
SOURCES & LIMITS
Sources, timing, and limits of interpretation
This issue covers October 5–11, 2026, and was backfilled on October 11. District policy, education events, and university initiatives are public rules, training arrangements, or project descriptions. They do not provide common learning or purchasing data.
MIT for America and RAISE describe support directions, not results at each school. LACOE’s count is an event measure; NAAIC’s 25 people are a course cohort. They are not comparable market reach. We did not test products or obtain district contracts.
