Digitally Literate
Enter your search text in the box above

Select a result to preview

Who Controls the Screen

Published: September 9, 2026 • 📧 Newsletter

Hi all, welcome back to Digitally Literate.

There are times when the AI news is breathless. This week felt like that, but for all of the wrong reasons. But even as things move fast, we need to think about why we’re allowing them to move fast and, more importantly, why we feel like we need to keep up.

This week, I started getting back to publishing on my main site:

As always, your support is valued. Reach out anytime at hello@wiobyrne.com, and subscribe if you haven’t already.

We cannot make policy fast enough

On September 2, New York City banned student-facing generative AI from 2-K through eighth grade, prohibited companion chatbots across all grades, and shut off AI in 38 previously approved programs. The Los Angeles Unified School District (LAUSD) moved in the same direction, temporarily blocking generative AI on student devices.

Two days later, Reps. Josh Gottheimer and Jay Obernolte announced the bipartisan AI LABS Act, which would make federal funding available for K-12 schools to establish AI labs and train teachers. The pitch is almost the mirror image of a ban. Schools need places where students can learn to use AI, question it, and understand when not to trust it.

What’s interesting to me is Gottheimer’s agenda, because it holds two opposite ideas at the same time. He wants to fund AI labs in schools while also backing the UNPLUGGED Act. The UNPLUGGED Act would use federal money to buy pouches and lock boxes that physically remove personal phones from classrooms. So we ban one device, and build a room for the other.

This sort of incoherent policy is not new in our educational spaces, as we label one set of screens good while the other is banned. I’m beginning to wonder if it’s more about who controls the screen. If the screen is owned and operated by the student, it’s banned and confiscated. If the screen is funded, owned, and operated by the district and authority, it’s supported.

Underneath them is the same harder question. Who gets to decide what this technology does, where it belongs, and how much power we hand over to it? The policies will keep changing. That question probably won’t.

We cannot learn fast enough

This week was equally breathless in terms of the models launched by the AI labs. First, Anthropic updated Fable and Mythos. Then came model enhancements from Meta and Google. OpenAI followed suit by releasing GPT-6 Astra.

OpenAI’s GPT-6 Astra provided the sharpest example of how quickly the ground is moving. Released September 3, Astra became the first OpenAI model to reach the company’s “Critical” level for cybersecurity capability.

OpenAI introduced its Preparedness Framework in 2023, and it serves as the company’s method for “tracking and preparing for advanced AI capabilities that could introduce new risks of severe harm.” “Critical” does not mean OpenAI has concluded the model itself is about to launch a cyberattack; it’s a label or threshold for what a model is capable of.

The model is no longer merely answering questions about cybersecurity. Given tools, time, and a target, it can carry out substantial parts of the investigative work itself. Put simply, Astra could look at a heavily protected computer system, discover a security hole nobody yet knows about, figure out how to exploit that hole to break in, and do much of that work without a person walking it through each step.

And the kicker...OpenAI only confirmed its capabilities after running additional tests shortly before release.

Maybe keeping up is the wrong goal

On September 4, the Council of Europe adopted a recommendation on AI literacy for its 46 member states. What caught my attention was not another call for schools to teach prompting or prepare students for an AI workforce.

It was the Council's recommendation to define AI not simply as a technology but as a sociotechnical phenomenon. Beyond understanding the systems themselves, we need to understand the people, institutions, policies, and social practices that shape their design and use. It asks what people need to understand in order to live and act in a society shaped by AI.

This sets the recommendation apart from other AI-literacy frameworks, which mainly focus on technical skills and reflect the Council of Europe’s long-standing conviction that education is one of the pillars of democracy. The recommendation includes 22 guiding principles organized around autonomy, privacy, equity, critical thinking, and social agency.

Instead of thinking about what models are available, who has access, and who controls them, the Council of Europe asks what capacities should remain with the person, no matter which system arrives next.

The Understory

Around 1817, a French engineer named Claudius Crozet began using a blackboard to teach mathematics at West Point. The blackboard itself wasn’t new, but West Point helped establish it as a serious classroom technology in the United States.

What made it powerful wasn’t that it displayed information; it made thinking visible. An instructor could pull mathematical reasoning out of a single cadet’s head and put it in front of the whole room, where the process could be inspected, corrected, and repeated. A problem that once lived on an individual student’s slate or sheet of paper could now be worked out in front of everyone, with each step visible.

The same logic ran through everything that surrounded the blackboard. Individual slates let teachers watch students produce work, then erase it for the next round of drill. The system allows one adult to govern hundreds of children through sequenced exercises and prescribed movements.

Schools have never simply adopted the technologies that make learning easier. They’ve repeatedly favored the technologies that make learning visible, repeatable, scalable, and governable. That’s the through-line. Adoption was never about ease. It was about control.

Technology can ultimately serve whatever purpose and use we define for it. The blackboard can make a private student's thinking public to the teacher. If we so choose, generative AI can do the reverse and pull the most important parts of a student’s thinking out of the room and into a private exchange the teacher never sees.

By the late nineteenth century, the blackboard was described as a mirror of the student’s mind. The teacher could see not only whether a student had reached the right answer, but how they’d gotten there. Two centuries later, we’re introducing a classroom technology that can do almost the opposite, if we frame it that way. The student can now move part of the work off the board entirely, asking questions, generating ideas, testing explanations, revising sentences, solving problems, sometimes producing the final answer in a conversation the teacher may never see.

The blackboard made thinking more visible to the institution. AI can make it less so.

See you next Wednesday. As always, my email is hello@wiobyrne.com.

Previous: DL 446 • Next: DL 448 • Archive: 📧 Newsletter

🕸️ Connected Concepts: