Utah drafted AI standards for 7th and 8th grade. What else changed this week?

·

Levitation Education

Three things worth knowing. Utah has a first draft of the digital literacy standards HB 218 set in motion, it contains a dedicated AI strand, and the board's timeline puts public comment in November. Washington's chatbot law exempts classroom tools in a sentence almost identical to Georgia's. And California's student data law, now readable in full, does bar ed-tech companies from training AI on student information.

Nothing new was signed anywhere this week. What changed is what can now be read.

Has Utah published its AI standards for 7th and 8th grade?

Not published, no. But a first draft exists, it is public, and it has an AI strand — further along than anything Utah had put out before.

The draft sits in the State Board of Education's packet for the Standards and Assessment Committee meeting of October 8, 2026, as agenda item 1.11, "ACTION: Digital Literacy Standards Update." It is headed "DIGITAL LITERACY — DRAFT 1," runs 24 pages, is watermarked DRAFT, and sets the effective school year as 2027-2028, grades 7-8.

This is the course HB 218 amended. The posted version is stamped May 2023, has six strands and no AI strand. The draft adds one:

STRAND 3 (ARTIFICIAL INTELLIGENCE (AI) LITERACY) Students will build foundational AI literacy by developing the knowledge and skills needed to understand, use, and responsibly interact with AI tools.

Underneath it are four standards: using AI "responsibly and effectively," including "recognizing when to use AI" and "how AI has contributed to a product, and communicating that clearly"; "the importance of transparency and documentation, both in their own use of AI tools and in verifying the authenticity of others' work"; comparing "machine learning, neural networks, generative, agentic" systems, with definitions of AI Bias and AI Hallucinations; and writing and refining prompts.

Two cautions. This is a draft, not a requirement: the board's timeline runs public comment in November 2026, then refinement and committee finalization, and full board approval in February 2027, for implementation in 2027-2028. And although it was listed as an ACTION item, Utah has posted no minutes for October 8, so whether the committee voted to release the draft is not yet on the record.

If you teach in Utah, the comment period is the thing to watch. It is the one moment when what these standards say is still open to argument.

Does Washington's new AI chatbot law apply to classroom tools?

Mostly not, but the exemption is narrow enough that a general-purpose classroom chatbot falls back inside the law.

Engrossed Substitute House Bill 2225 is Chapter 168, Laws of 2026, signed by Gov. Bob Ferguson on March 24, 2026 and effective January 1, 2027. Final passage in the House was 74-21 on March 11, after the Senate passed it 43-5. Section 2(1)(a) defines an "AI companion chatbot" as

an artificial intelligence system with a natural language interface that provides adaptive, human-like responses to user inputs, including by exhibiting anthropomorphic features, and is able to sustain a relationship across multiple interactions.

All three parts must be true at once. The school exemption is at Section 2(1)(b)(iv):

Narrowly tailored educational tools used in school or instructional settings that are designed solely to support specific, curriculum-aligned learning objectives and do not provide open-ended conversational companionship.

Narrowly tailored, instructional, solely curriculum-aligned, not open-ended. A chatbot that will talk about anything does not clear that bar. Outside that sentence, the words school, student, teacher and classroom appear nowhere in the act.

And that sentence is not Washington's invention. Georgia's SB 540, at O.C.G.A. § 39-5-6(a)(1)(B)(v), exempts "A narrowly tailored educational tool designed solely to support specific curriculum aligned learning objectives" designed "not to provide open ended conversational companionship." Washington says the same thing in the same order, adding only "used in school or instructional settings." Two legislatures on opposite coasts, one sentence: model language moving between states, and a repeated four-part test for buyers. Georgia's bites July 1, 2027; Washington's first, on January 1, 2027.

Can ed-tech companies train AI on student work in California?

No. AB 1159 is Chapter 182, signed September 10, 2026, and the prohibition is explicit. An operator shall not knowingly

Use covered information, including persistent unique identifiers, created or gathered by the operator's site, service, or application to train a generative artificial intelligence system or service or develop an artificial intelligence system.

That language appears twice — at Business and Professions Code § 22584(b)(5) for K-12, and at § 22586(b)(5) for preschool. "Covered information" is broad: anything personally identifiable created by a pupil or gathered about one.

Three details matter. It binds operators — companies running a service "designed or marketed for K-12 school purposes" — and expressly not schools, districts or the state. It does not stop a company improving its own product, or using pupil data "for adaptive learning or customized pupil learning purposes." And it has teeth: a new § 22585 lets a harmed pupil or parent sue for "the greater of actual damages or five hundred dollars ($500) per plaintiff, per violation," plus punitive damages and fees. The same act extends the scheme to colleges from July 1, 2027.

Separately, AB 1792 is Chapter 273, signed September 18, 2026. It adds Education Code § 51900.3 and does mention AI — directing the Instructional Quality Commission to consider including, in the next health framework revision, instruction on "deepfakes, online grooming, sextortion, stalking, and the misuse of generative artificial intelligence." Consider, not require, and it sits inside sexual health instruction. It is not an AI literacy mandate, whatever a tracker may say.

What does Florida's new college AI rule require, and when?

It requires every Florida College System institution's board of trustees to adopt AI policies, and it takes effect on October 20, 2026 — next Tuesday. Rule 6A-14.0719 was filed for adoption on September 30. It reaches colleges, not school districts.

Policies must require "parents of enrolled minors to be notified if their child will be directly using an artificial intelligence instructional tool," which reaches dual-enrolled teenagers; must prohibit AI on graded work unless the instructor permits it; and must adopt an "Approved A.I. literacy definition or framework," pointing to the US Department of Labor's AI Literacy Framework.

Florida's companion-AI language reappears as a definitional exclusion rather than a ban: a tool is not an "artificial intelligence instructional tool" at all if it is designed, marketed or configured to "Meet a student's social needs," "Simulate friendship, companionship, or an emotional relationship with a student," or employ "relationship-building or anthropomorphic design features" to keep a student interacting. Only instructional tools can be approved, so the effect is the same.

Do Oregon and Idaho's chatbot laws mention schools at all?

Between them, once — and it is not a carve-out.

Idaho SB 1297, the Conversational AI Safety Act, is Chapter 249, Laws of 2026, signed March 31 and effective July 1, 2027. Oregon SB 1546 is Chapter 85, Oregon Laws 2026, signed the same day, effective January 1, 2027. Idaho regulates a "conversational AI service"; Oregon an "artificial intelligence companion." Neither contains the words school, student, teacher or curriculum anywhere. Oregon's single use of "education" is a bare item in a list of purposes, excluding software that "operates solely for the purpose of" any of them — one word gated by the word solely.

So a classroom tool in those two states has to fall outside the general definition or into a general exclusion, not an education one. Idaho's likeliest routes are its exclusions for outputs "relating to a narrow and discrete topic" and for services reached only under "a commercial agreement, enterprise contract, or similar business arrangement." Idaho now sits on both sides of this beat: SB 1227 builds AI literacy into K-12, SB 1297 never mentions it.

What did not change

Florida's K-12 rule is still not in the code. Rule 6A-1.0957 shows an effective date of August 22, 2023, the newest filing against it being the change notice of September 18, 2026. The board approved the amendment but it has not been adopted, so the July 1, 2027 compliance date rests on that text, not the code.

New York's chatbot bill is still sitting. S 9051-B passed the Senate 60-0 on June 4 and the Assembly on June 5. Nothing has happened since, and it has not been delivered to the Governor.

The federal bills have not moved. H.R. 8747's last action remains July 21, 2026, "Ordered to be Reported (Amended) by the Yeas and Nays: 18 - 15," with no committee report filed; S. 5225 went to committee on August 4. The House Education and Workforce Committee has held no markup since July 21.

North Carolina is still in conference. H 301, "Social Media & AI Safety," has not moved since conferees were appointed on June 23 and 24, and S 864 has sat in Senate Rules since April 29. Both chambers reconvene October 26, making North Carolina the likeliest source of the next enactment here.

And no other state has copied Florida's companion-AI prohibition into its own K-12 rules.

What does this mean if you are choosing a course now?

Two things are converging, and neither is curriculum quality. The chatbot statutes are converging on whether a tool holds an open-ended relationship with a child: nine states are verified enacted against their own legislatures, five more unchecked, so we publish no count — but each either carves out curriculum-bound tools narrowly, as Georgia and Washington do, or says nothing about education and leaves a general narrow-topic exclusion to do the work. The privacy statutes are converging on who holds the data, and California's answer, with a private right of action behind it, is the strictest yet.

Both point at the same tool: one that works from fixed course material, answers within it, holds no personal data, and neither remembers a learner's life nor asks how they are feeling. If a vendor cannot say plainly where they sit on both, press on that.

Where does ZEROTH stand?

ZEROTH is an AI ethics and literacy course for ages 13 and up that runs as a private game in your own Google account. Its Almagest answers from seven course files and nothing else, holds no memory of a learner's personal life, and never asks about feelings. Because the course runs in your own Google account, Levitation holds no learner data to train anything on. ZEROTH is independent curriculum, not accredited, not state-approved, awarding no credit; it is written to be aligned to the AI literacy language in Utah's HB 218 and Idaho's SB 1227. A crosswalk is available on request, and we are rebuilding it against Utah's draft strand.

This is not legal advice. Statuses change quickly; this roundup reflects what the states' own legislatures, boards and registers showed on October 14, 2026. Utah's digital literacy standards are a draft and may change before the board votes. If a requirement might apply to your school, district or co-op, check your state's current law and take advice.

FROM READING TO DOING

World 1 of ZEROTH is free

An evening's work, no email address, no account with us. It is the quickest way to find out whether this is the kind of AI teaching you want in your house or your classroom.

Levitation Education

Published by Levitation Automation.

ZEROTH and the Operator turtle are marks of Levitation Automation. Independent curriculum. Not affiliated with or endorsed by Google or any other tool provider named in the course. Not accredited; awards no credit.

© 2026 Levitation Automation.