A Unanimous Vote Is a Message, Not Just a Bill
On June 24, California's AB 2148 was sent to Governor Gavin Newsom. The bill does something deceptively simple: it specifies that public school employees serving as teachers must be natural persons, not artificial intelligence systems. What makes it remarkable is not the text but the margin. It cleared the Assembly 76 to 0 in May and the Senate 38 to 0 in June. In a legislature that agrees on almost nothing unanimously, lawmakers agreed that a machine should not be the teacher of record. That consensus is the story.
We have spent two years watching the education technology industry promise AI tutors, AI graders, and AI teaching assistants while regulators stayed vague and advisory. That phase is ending. A unanimous, bipartisan vote to put the word "human" into statute is a different kind of signal than a non-binding guidance memo. It tells vendors, and the executives who buy from them, that the boundary between AI assisting a classroom and AI being the classroom is about to become a matter of law rather than marketing taste.
The Same Logic Is Climbing Into Higher Education
California is not stopping at K-12. A companion bill, SB 928, would mandate that California State University instructors be human and not AI, and it has already cleared the Senate 37 to 0 and passed an Assembly higher education committee. Alongside it, AB 2392 would require the community colleges, CSU, and, by request, the University of California to convene working groups on generative AI procurement standards and training before deploying systems to students, faculty, or staff. The pattern is coherent: protect the human role, then gate the procurement.
For anyone selling into higher education, AB 2392 is the clause to watch. Procurement working groups and pre-deployment standards are exactly the kind of friction that lengthens enterprise sales cycles and raises the compliance bar. The implicit message to vendors is that institutions will increasingly be required to vet AI systems through formal governance before adoption, not after a pilot goes sideways. That is a structural change in how AI gets into a university, and it favors providers who can document training, oversight, and data handling up front.
One State, but a National Wave
California is the loudest example, not the only one. MultiState is tracking 134 bills across 31 states in the 2026 legislative session, all addressing AI in education. The bills cluster into three themes that should be familiar to any technology leader: student data privacy, human oversight of high-stakes decisions, and AI literacy or curriculum requirements. This is no longer a handful of cautious memos. It is a coordinated, if uneven, legislative movement to define where AI belongs in schools and where it does not.
The specifics vary by statehouse but rhyme. Idaho's SB 1227, already enacted, requires a statewide framework for AI in K-12 schools, sets data privacy requirements for AI tools, mandates educator training, and prohibits AI from replacing human teachers. California's AB 1159 moves to bar student data from being used to train AI models. Others, in states from Oklahoma to Maryland, restrict AI from making high-stakes decisions about students without human review. The throughline is accountability: a person must remain answerable for what happens to a child.
Graduation Requirements Are the Other Half of the Story
While one set of bills restricts AI, another mandates teaching about it. Georgia's SB 179 makes computer science, including AI, a high school graduation requirement beginning in 2031 to 2032. Alabama's HB 329 requires students to complete a computer science course that includes AI instruction to graduate. New Jersey is pushing AI concepts and ethics into the K-12 curriculum and asking public universities to stand up AI certificate and degree programs. The same legislatures wary of AI replacing teachers are insisting students learn to use it well.
We see no contradiction in that pairing, and executives should not either. The emerging policy posture treats AI as a subject and a tool, not a substitute for the educator. That distinction creates real demand: curriculum, professional development, assessment, and literacy products all sit on the right side of where the law is heading. The vendors who will struggle are those whose pitch quietly assumed AI would do the teaching. The states have now told them, with unusual unanimity, that the answer is no.
What This Means for the People Buying Edtech
For CIOs, CTOs, and the administrators who sign these contracts, the regulatory direction is now clear enough to plan around. AI that augments a human educator, surfaces insight, drafts materials, or personalizes practice under teacher control is on safe ground and increasingly mandated as a skill. AI positioned to replace the educator, make consequential decisions autonomously, or train on student data is walking into a thickening wall of statute. Procurement decisions made today should be stress-tested against that trajectory, not against the permissive vacuum of 2024.
The strategic read is straightforward. Buy AI that keeps a human accountable and documents its data handling, training, and oversight. Demand procurement-ready compliance from vendors before deployment, because California is about to require exactly that. And treat the 134-bill wave not as noise but as a forecast: the legal definition of who may teach, and what may be done with student data, is being written right now, and it is being written in favor of the human in the room.
