November is fully booked! Next classes blast off the week of Mon, Dec 7.

Reserve your spot
What DC's Stoplight AI Rulebook Actually Says, Band by Band
Table of Contents

What DC's Stoplight AI Rulebook Actually Says, Band by Band

DC's stoplight AI rulebook sorts school AI uses into green, yellow and red. Here is what each band allows, what it bans, and what every parent should ask first.

Most school AI policies tell you what software is approved. DC’s stoplight AI rulebook does something rarer: it names decisions about children that AI may never touch. Released September 1, 2026 by the District’s Office of the State Superintendent of Education for the 2026–27 school year, the guidance sorts uses into green, yellow and red. The red band includes student discipline decisions, individualised education plans for students with disabilities, teacher performance evaluations, and physical surveillance. That list, not the approved-tools list, is the useful part.

Key Takeaways

  • The red band prohibits AI from four specific decision types, including discipline and IEP drafting. Most district policies never name a prohibited decision at all.
  • Grading sits in yellow, not green. State Superintendent Antoinette Mitchell: “for the most part, it feels like grading should be something that a teacher does based on their own judgment.”
  • The guidance is not mandatory. Schools requested it anyway, which tells you how little written clarity existed before September 2026.
  • Green uses are almost entirely teacher-facing and administrative: lesson planning, data analysis, scheduling, drafting parent messages.
  • A stoplight framework is only as strong as its red band. If your district’s document has no red list, that absence is the policy.

What a stoplight AI rulebook is, and why DC built one

A stoplight AI rulebook is a policy document that sorts permitted AI uses into three tiers rather than issuing a single yes or no. Green means go. Yellow means proceed with mandatory human oversight. Red means stop.

The format matters because the alternative, a flat ban or a flat approval, collapses very different activities into one rule. Using an AI to draft a parent newsletter and using an AI to decide whether a ten-year-old gets suspended are not the same act. A single-answer policy has to treat them the same. A tiered one does not.

DC’s guidance covers the 2026–27 school year and was reported by WTOP on September 1, 2026. Mitchell framed the yellow band in the clearest sentence any education official produced that month: “Yes, you probably can use AI in these areas, but with a lot of human oversight and also, humans making the final decisions.”

The three bands, with the actual contents

Here is what WTOP reported falls where. I have not paraphrased the categories into friendlier language, because the specificity is the point.

BandWhat it coversWhy it lands there
GreenDrafting lesson plans; analysing assessment or attendance data; scheduling; drafting messages to parentsTeacher-facing, reversible, no consequential judgment about an individual child
YellowAssisting in reviewing or grading student work; monitoring digital activity on school-issued devicesTouches a child’s record or privacy; requires heavy human oversight and a human making the final call
RedStudent discipline decisions; individualised education plans for students with disabilities; teacher performance evaluations; physical surveillanceConsequential, hard to appeal, and in two cases legally protected

Two things stand out. First, every green use is something a teacher does to paperwork, not to a student. Second, the red band is not about capability. Nobody is arguing an AI could not draft an IEP. The argument is that it should not, because the document carries legal weight and the reasoning behind it has to belong to a person who can be questioned.

The guidance also states that schools must ensure “staff only use those platforms and tools that are approved by their district,” which puts tool vetting on districts rather than individual teachers.

Why grading is yellow and not green

This is the band most parents will care about, and the reasoning is worth understanding.

Grading is a prediction about what a child knows. The IES practice guide Organizing Instruction and Study to Improve Student Learning (Pashler et al., 2007) rates re-exposure quizzing as having strong evidence for producing durable learning, precisely because retrieval itself strengthens memory. In other words, the act of being assessed is instructional, not just administrative. If a machine does the reading and a teacher signs the number, the teacher loses the diagnostic information that grading generates: which misconception showed up in eleven papers, which kid suddenly understood fractions.

That is a pedagogical argument, not a privacy one. Mitchell’s phrasing lands on judgment: grading “should be something that a teacher does based on their own judgment.”

Compare New York City, which went further. Under the moratorium announced September 2, 2026, NYC teachers may use AI for instructional planning and operational tasks but are barred outright from using AI tools to grade student assignments. Same concern, harder rule. Our breakdown of the three school AI policy models now in play covers how differently districts answered the same question that month.

What “not mandatory” actually means for your kid

DC’s guidance is advisory. A school can decline to follow it and break no rule.

That sounds like a weakness, and partly it is. But non-mandatory guidance does real work in one specific way: it creates a document a parent can point at. Before September 2026, a parent asking “is my principal allowed to run disciplinary decisions through an AI?” had nothing to cite. Now there is a state-superintendent document saying no.

Compare the enforceable route. On September 9, 2026, the American Federation of Teachers, the United Federation of Teachers and Microsoft announced a National AI Safety & Privacy Standard for schools, effective November 1, 2026. Its protections are specific: “Student and educator data will not be used to train AI models, sold or repurposed”; “Schools will maintain control over how data is used, retained and deleted”; “Students can never be tracked”; “AI cannot make decisions without human oversight.” Districts can write these terms into their Microsoft customer agreements, which makes them contractually binding. We go through the agreement clause by clause in the Microsoft and AFT AI safety and privacy standard. UFT President Michael Mulgrew described the result as districts gaining power to end agreements and seek damages from bad actors who break the rules.

That is the difference between a stoplight and a contract. The stoplight tells a teacher what good practice looks like. The contract gives a district a remedy.

The gap the stoplight leaves open: your kid

Read the bands again and notice who is missing. Green, yellow and red all describe adult conduct. Nothing in the framework tells a seventh grader whether she may ask a chatbot to outline her history essay.

That is not an oversight so much as a division of labour. Institutional governance and student conduct are different documents in most districts, which is why a family can read an AI policy carefully and still have no idea what their own child is allowed to do. Florida took the opposite tack: its State Board of Education voted September 16, 2026 to require affirmative parental consent before a student interacts directly with an AI tool, with a comparable non-AI alternative for families who decline and a July 1, 2027 implementation deadline. That rule says almost nothing about how teachers may use AI, and a great deal about students.

So the two documents answer opposite halves of the question. DC constrains the institution and leaves student use to local practice. Florida constrains student access and leaves institutional use largely alone. A family wanting the full picture needs both kinds of document, and most districts have published neither.

Pew Research gives a sense of the stakes on the student side: 26% of US teens had used ChatGPT for schoolwork in its survey of 1,391 teens fielded September 18 to October 10, 2024, double the 13% recorded in 2023. Teens drew their own lines, too. In that survey 54% said using it to research a new topic was acceptable, 29% said the same for solving maths problems, and only 18% thought writing essays with it was acceptable while 42% said it was not.

How to read your own district’s AI document using this frame

Find the red list first, not the tool list

Open the document and search for words like “prohibited,” “shall not,” “never.” If the only prohibitions concern students cheating, the policy is an academic integrity memo wearing an AI costume. A real AI governance document constrains the institution, not just the kids.

Check whether discipline is named

Of DC’s four red items, this is the one most likely to affect your child directly. Behaviour-flagging software already exists in many districts under the heading of safety monitoring. Ask in writing: does any automated system contribute to disciplinary referrals at this school, and who reviews its output?

Check whether IEPs are named

If your child has an IEP or a 504 plan, this matters more than anything else in the document, more even than the question of what a district AI block actually blocks. An AI-drafted IEP can read fluently and still miss the specific accommodation that works for your kid, because the model has no history with your kid. DC put this in red. Ask whether yours did.

Ask what happens in the yellow band when a teacher disagrees with the AI

Yellow means human oversight, which only means something if the human can override. Ask: if an AI grading assistant marks an essay a 72 and the teacher thinks it is an 88, which number enters the gradebook, and does anyone record the disagreement? Districts that cannot answer this have not implemented oversight. They have implemented a rubber stamp.

Ask which tools are on the approved list and who approved them

DC’s guidance pushes approval to the district level. That is good governance and a real workload. Instruction Partners reviewed 20 classroom AI tools across 16 school systems in its AI Learning Tour, published September 14, 2026, and found the largest risks came from general-purpose chatbots rather than purpose-built instructional tools. Co-founder Emily Freitag’s summary: “Not all AI is created equal… We need nuanced decisionmaking about where it can offer real value.” Ask whether your district’s approval process distinguishes the two categories.

What not to do: do not treat green as “safe for my child”

Green in DC’s framework means safe for a teacher’s workflow. Drafting a parent message with AI is green, and it is also how a factual error about your child reaches your inbox with a school letterhead on it. Green is a judgment about institutional risk, not a guarantee of accuracy.

What to Watch For Over the Next 3 Months

  • Week 4: Watch for the first district to publish a red band that copies DC’s four items. Copying is how a norm becomes a standard, and the Microsoft–AFT standard took effect November 1, 2026, which gives districts a contractual hook for the same protections.
  • Month 2 red flags: A revised policy that quietly moves grading from yellow to green. An approved-tools list with no named reviewer. A document that prohibits student use but places no limit on institutional use.
  • Month 3 self-check: Email your principal one question: “Which decisions about my child is AI not permitted to make at this school?” A specific answer means the policy is real. A reassuring answer with no list means it is not.

Frequently Asked Questions

Is DC’s stoplight AI rulebook legally binding on schools?

No. WTOP reported the guidance is not mandatory, though schools and administrators had requested it. Its practical value is as a citable reference point for parents and as a template other districts can adopt.

Can my child’s school use AI to monitor what they do on a school laptop?

In DC’s framework, monitoring digital activity on school-issued devices sits in the yellow band, meaning it is permitted with substantial human oversight and human final decisions. Physical surveillance is red. Ask your district which monitoring products are in use and who reviews alerts.

Does the stoplight apply to students or to teachers?

Primarily to educators and administrators. Every green item is a teacher or administrative task. DC’s guidance governs how adults in the building may use AI about children, which is a different question from whether students may use AI on homework.

My district has no AI policy at all. What do I ask for?

Ask for three things in writing: a list of decisions AI may not make, a list of approved tools with a named approver, and a statement on whether student data may be used to train vendor models. The AFT–Microsoft standard shows all three are achievable contract terms.

Should I be more worried about AI grading or AI monitoring?

Monitoring, in most cases. A disputed grade has an appeals path most families already know how to use. A behaviour flag often has none, and the parent may never learn it was generated. Ask about appeals for both.


About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.


Sources

  1. WTOP News. (2026, September 1). “No, robots won’t be teaching your kids: DC releases new AI guidelines for teachers.” https://wtop.com/dc/2026/09/no-robots-wont-be-teaching-your-kids-dc-releases-new-ai-guidelines-for-teachers/
  2. Microsoft News Center. (2026, September 9). “AFT, UFT and Microsoft announce National AI Safety & Privacy Standard for schools.” https://news.microsoft.com/source/2026/09/09/aft-uft-and-microsoft-announce-national-ai-safety-privacy-standard-for-schools-to-protect-students-families-and-educators/
  3. Pashler, H., Bain, P., Bottge, B., Graesser, A., Koedinger, K., McDaniel, M., & Metcalfe, J. (2007). “Organizing Instruction and Study to Improve Student Learning.” IES Practice Guide, National Center for Education Evaluation. https://ies.ed.gov/ncee/wwc/practiceguide/1
  4. Education Week. (2026, September 14). “New Project Identifies Strengths and Weaknesses of a Collection of AI Learning Tools.” https://www.edweek.org/technology/new-project-identifies-strengths-and-weaknesses-of-a-collection-of-ai-learning-tools/2026/09
  5. Hoodline. (2026, September 2). “NYC Bans AI for 600,000 Students Through 8th Grade.” https://hoodline.com/2026/09/nyc-bans-ai-for-600-000-students-through-8th-grade-cellphone-style-crackdown/
  6. UNESCO. (2023). “Guidance for Generative AI in Education and Research.” https://www.unesco.org/en/articles/guidance-generative-ai-education-and-research
  7. Pew Research Center. (2025, January 15). “About a Quarter of U.S. Teens Have Used ChatGPT for Schoolwork.” https://www.pewresearch.org/short-reads/2025/01/15/about-a-quarter-of-us-teens-have-used-chatgpt-for-schoolwork-double-the-share-in-2023/
  8. Central Florida Public Media. (2026, September 16). “Florida Approves Statewide AI Rules for K-12 Schools and Colleges.” https://www.cfpublic.org/education/2026-09-16/florida-approves-statewide-ai-rules-for-k-12-schools-and-colleges
Ricky Flores
Written by Ricky Flores

Founder of HiWave Makers and electrical engineer with 15+ years working on projects with Apple, Samsung, Texas Instruments, and other Fortune 500 companies. He writes about how kids learn to build, think, and create in a tech-driven world.