This school year, whether you're allowed to use AI depends a lot on where you're sitting.
New York City Public Schools, the largest school system in the country, put a one-year moratorium on student-facing generative AI for students from 2-K through eighth grade, with high schoolers limited to approved, vetted programs. UC Berkeley School of Law banned students from using generative AI on any graded work and during exams. Across the country, Columbia Law School went the other way, allowing students to use AI as a learning aid (summarizing notes, generating practice hypotheticals) while barring it from anything they submit. Ohio required every school district to adopt an AI policy by July 1, 2026, and Idaho, Maryland, Oklahoma, and Virginia all passed education-specific AI measures this year.
Meanwhile, 98 high school students from all 50 states spent three days in Boston this summer drafting their own national framework, the STUDENTS FIRST Act of 2026.
That's a lot of different answers to the same question. So here's ours: the question isn't really "AI, yes or no?" It's "what is this AI doing to the student's thinking?"
The cautious policies make sense (really)
We want to be fair here, because the schools pumping the brakes aren't wrong to worry. A chatbot that will write a finished essay in ten seconds is a genuine threat to learning. If the tool does the thinking, the student doesn't, and no grade can fix that. Younger students are still building the basic skills that AI can so easily skip over. And schools have every right to want time to study what this technology actually does in a classroom before handing it to hundreds of thousands of kids. New York framed its pause in exactly those terms.
Thea has been anti-cheating from day one. We don't want students outsourcing their education any more than teachers do.
But "AI" isn't one thing
Here's where a lot of policies get blurry. A tool that writes your essay and a tool that quizzes you on your own class notes are both technically "AI." But they have opposite effects on learning. One does the thinking for you. The other makes you do more thinking than you would have on your own.
The research is starting to point in the same direction. A June 2026 review of 22 studies on generative AI risks in K-12 education suggested schools can reduce those risks by using AI to offer hints rather than complete answers, assessing how students work and not just what they turn in, and teaching students how to evaluate what AI produces.
In other words: the problem isn't AI in the room. It's AI that hands over the answer.
The students got a lot right
Some of the most thoughtful ideas this year came from the students themselves. The STUDENTS FIRST Act calls for:
- AI-literacy instruction as soon as students start using digital devices in school.
- Using AI as a supplement to learning, never as a way around it.
- A clear way for students to appeal accusations of wrongful AI use.
- The right for high school students to decline AI use and receive an alternative assignment.
That appeals process deserves a spotlight. As schools tighten their rules, more students are being accused of using AI on work they wrote themselves. Any policy strict enough to punish students should be fair enough to let them defend themselves.
What we think good AI policy looks like
We're a study platform, not a school board, so take this as a wishlist from people who think about learning all day:
- Judge tools by what they do to thinking. Does the tool produce the work, or does it make the student produce it? That one question sorts most tools pretty quickly.
- Teach AI literacy early. Students are going to meet AI either way. Better they learn how it works (and where it's wrong) in a classroom.
- Keep teachers in the driver's seat. The best tools are built from a teacher's own course material, aligned to what's actually being taught and tested.
- Build in fairness. Clear rules per class, and a real process when a student is accused.
- Revisit often. This technology changes every semester. Policies written in 2026 shouldn't be frozen in 2026.
Where Thea stands
Thea was built on the idea that AI should ask the questions, not answer them. You upload your own notes, slides, or readings, and Thea turns them into practice questions, flashcards, and study guides. You still do the thinking. You just do more of it, more effectively. We've designed Thea alongside educators, not around them.
And when educators bring that approach into their classrooms, the results speak for themselves. In one workforce certification program, faculty uploaded their exact course materials to Thea and students' Network+ pass rate went from 46% to 100% in a single semester. (Read the full case study.)
If you're a student right now
While the policies keep shifting, here's how to stay on the right side of them:
- Check every syllabus. AI rules can change class to class, even within the same school.
- When in doubt, ask. "Is it okay if I use AI to quiz myself on the readings?" is a question you can ask.
- Keep your receipts. Draft in tools with version history so you can show your process if you ever need to.
- Use AI to study, not to submit. If AI is making you think harder, you're probably doing it right. If it's thinking for you, stop.
The schools, the states, and the students are all working toward the same thing: making sure people actually learn. We think the right kind of AI can help with that. The wrong kind can't. The best policies will be the ones that can tell the difference.
Want AI that makes you think instead of thinking for you? Try Thea for free.
