What She Did Not Ban

On Wednesday, at the National Press Club, Randi Weingarten gave the most detailed speech a major American teachers union president has yet given about technology in classrooms. The title of the address was a sentence on its own: Devices Down, Eyes Up, Hands-On: 10 Points to Boost Teaching and Learning in the AI Era.¹

The press coverage moved fast. The American Federation of Teachers wants a complete ban on screens, including online assessments, for pre-K through second grade, with exceptions only for students with disabilities.² An end to all student-facing artificial intelligence, including digital tutors, in elementary school.² A ban on "companion chatbots," meaning any AI designed to simulate a human relationship, for anyone under sixteen.² A tax on the earnings of large tech companies to pay for the disruption they have caused.³ A "gold standard" for AI safety and privacy in K-12 schools that Microsoft, OpenAI, and Anthropic have, according to Weingarten, agreed to in principle.³

By the end of the day, the headlines were a single word: ban.

Read more carefully, the speech was not a ban on AI in schools. It was a much more careful line.

The Word Was "Companion"

Weingarten was specific about which AI she wanted out of children's hands and which AI she did not. The first category she named was "companion chatbots," programs that simulate friendship, advice, romance, or coaching as though they were a person.² Her concern was not that a child might use a calculator. It was that a child might form a primary relationship with a piece of software that has no obligation to her well-being.

The second category was "student-facing" AI in elementary school, tools that interact directly with a seven-year-old without a teacher in the loop.² Again, the line was not about whether AI exists in the building. It was about whether the youngest students are sent into it alone.

What Weingarten conspicuously did not ban was teacher-facing AI, AI literacy in older grades, or the use of these tools in the kinds of project-based, hands-on learning her ten-point plan also explicitly called for.¹ The AFT's own National Academy for AI Instruction, a training hub designed and run by educators, continues its work.³

The shape of that line is the part of the speech that will matter five years from now. Weingarten quoted a survey of three thousand teachers in which eighty-eight percent said their students' attention spans were getting shorter.² She cited research from the neuroscientist Jared Cooney Horvath, who found that state expansion of ed tech correlated with declining educational progress.³ "Intentional or not," she said, "all this tech has been a huge experiment on kids, and experiments can go wrong."²

What the Future Will Have to Decide

The choice in front of American schools right now is not yes-AI or no-AI. It is about which AI, for whom, with what kind of adult in the room. Weingarten was, in her careful way, modeling the question. A six-year-old should not be alone with a screen for hours. A fifteen-year-old probably should not be talking to a synthetic friend at midnight. A high school junior writing a research paper probably should be learning to work alongside the tool she will use for the rest of her working life.

The temptation, on hearing a speech like Wednesday's, is to read it as a referendum on AI. It is not. It is a referendum on relationships. Who is sitting next to the student when the work is hard. What the software is being asked to be. Whether the screen is replacing a person or supporting one.

That is a future worth building toward. It also implies a different burden of evidence than ed tech has historically been asked to meet. If the worry is that screens have replaced something irreplaceable in early childhood, then the case for AI in any classroom rests on its ability to make the human work more visible, not less. The drafts a student kept. The pause she took before answering. The question she asked her teacher because the tool taught her how to ask better questions.

This is the part Koan has been quietly built for. Not as a companion. Not as a substitute. As a way of preserving the trail of thinking that good teachers have always cared about and rarely had the time to see. The revisions, the pauses, the breakthroughs. The evidence that a sixteen-year-old is doing the work, not outsourcing it. The signal a teacher can use to sit down next to the right student tomorrow.

Weingarten's plan is not the end of AI in American schools. It is the beginning of a more honest conversation about what AI is allowed to be in a child's life. If the answer is "a tool that lets adults see learning more clearly," the future of education will look very different from the one the headlines were predicting on Wednesday afternoon.

If the line is not between AI and no AI, but between AI as a companion and AI as a witness, which side of that line is your school on?