Start with a set of dates.

September 8–11: UNESCO's "Digital Learning Week" was held at its Paris headquarters, themed "Education in the Age of AI." A team from Tsinghua University stood on that stage with an education agent project called OpenMAIC.

September 23: In Beijing, at a State Council Information Office press conference on the "15th Five-Year Plan," China's Ministry of Education called for "advancing the digital-intelligent transformation of education and strengthening AI literacy among teachers and students."

The same day, a report from the "2026 AI Large Model CEO Roundtable" proposed a five-layer AI literacy framework: "know, use, discern, create, be responsible."

September 24: At the Yunqi Conference in Hangzhou, TAL Education's "Jiuzhang Teacher" announced it had launched over 170 teaching agents, with more than 300,000 registered users.

September 24: Across the ocean, at its Connect conference, Meta unveiled "Muse Charm" — a small device that clips onto a keyring and is built to talk with its AI assistant, Muse — and said it plans to bring Muse into smart glasses.

September 25: Microsoft released a new "super app" version of Copilot, adding three capabilities: Home, Code, and Autopilot.

September 28: The Hong Kong Education Bureau's first AI-education teacher-training session began.

In a single month, from Paris to Beijing, from Hangzhou to Silicon Valley, education AI and consumer AI pressed the accelerator almost simultaneously. This is no coincidence.

This article tries to answer one question: when so many moves happen at once, what exactly are they competing over?

My assessment: they are fighting on two fronts at the same time — a battle over standards, and a battle over gateways.


1. China Is "Setting Standards": A Chain from Policy to Classroom

First front. China's moves share one trait: they answer questions like "what counts as AI literacy, how do we teach it, and to what level."

Link one: Policy sets the tone.

At the September 23 press conference, the Ministry of Education's phrase was "advancing the digital-intelligent transformation of education" — note "digital-intelligent," not merely "digital." The two characters point at the AI layer. The emphasis that day: "strengthening AI literacy among teachers and students."

The weight of that sentence: it elevates "AI literacy" from an optional bonus to a basic capability the education system must deliver.

Link two: A literacy framework lands.

Around the same period, a report from the "2026 AI Large Model CEO Roundtable" proposed a five-layer AI literacy framework: "know, use, discern, create, be responsible."

These five words are a progressive path: first "know" what AI is; then learn to "use" it; then "discern" (judge whether AI is right); then "create"; and finally "responsibility" (understand who bears the consequences of using AI).

One number in the report is worth pondering: among surveyed primary-school students, 97.9% had at least heard of AI, and 74.8% had already encountered AI in daily study. Yet the report's phrasing was "high awareness, shallow understanding."

In plain terms: kids all know AI, but no one is teaching them to truly understand it. That is exactly the gap the framework aims to close.

Link three: The standard goes international.

On September 8–11, at UNESCO's Digital Learning Week, the MAIC team from Tsinghua University's School of Education used a 90-minute presentation to bring OpenMAIC to international peers.

The significance: China's practice of "how AI enters education" has begun to enter the international conversation.

Link four: It reaches the classroom.

On September 28, the Hong Kong Education Bureau's first AI-education teacher-training session began (a repeat session was scheduled for October 5). This looks smallest yet matters most — standards ultimately land on teachers. If teachers can't use it, the framework stays in a document forever.

Connect the four links and you see a complete chain: policy tone → framework design → international dialogue → teacher training. From "should we do this" all the way to "how do teachers teach it."

That is the first layer of the "battle over standards": whoever defines "competent AI literacy" defines the starting line for the next generation.


2. Industry Is "Setting the Landscape": A Highly Concentrated Market

As standards form, the market is rapidly converging too.

First, one number: 79.4%.

In the first half of 2026, four brands — Zuoyebang, Xueersi, iFLYTEK, and Yuanfudao — together held 79.4% of unit-sales share in the learning-machine market, up 11.6 percentage points year on year.

(A note on scope: this is unit-sales share, for the period H1 2026. A separate, differently-scoped figure exists — in Q1 2026, six major learning-tablet brands together held 87.5%. The two should not be conflated.)

What does 79.4% mean? It means the market is already highly concentrated: four players take nearly 80%, and everyone else splits the remaining 20%.

Now a case: Jiuzhang Teacher.

On September 24 at the Yunqi Conference, TAL Education's "Jiuzhang AiXue Teacher Edition" (abbreviated "Jiuzhang Teacher") announced: over 170 teaching agents launched, covering the full teaching cycle — lesson prep, question-setting, homework grading, learning diagnostics, and parent communication — with more than 300,000 registered users.

Its positioning was stated clearly: "the professional teaching-research group for teachers in the AI era." It does not face students directly; it serves teachers.

Together, these two numbers reveal two features of China's education AI:

  1. The market is highly concentrated. The top four taking 80% signals a shift from "land-grab" to "consolidation of the top."
  2. Product form has shifted to "agents." From "an app" to "a cluster of agents covering the full teaching cycle."

The market also repeats a narrative — "AI learning machines to exceed 100 billion yuan in scale in 2026." To be clear: this is a forecast, not a fact. And the "100 billion" scope is usually the education smart-hardware category as a whole (including AI learning machines), not AI learning machines alone.

So the essence of the second front: it is not one product winning — it is an ecosystem taking shape.


3. Global Giants Are "Seizing Gateways": AI Is Becoming an Operating System

If China's moves concentrate on "standards" and "ecosystem," the global giants' keyword is another one: gateway.

First: Microsoft's Copilot "super app."

On September 25, Microsoft released a new Copilot adding three capabilities: Home, Code, and Autopilot.

Microsoft's positioning was blunt: to extend the AI assistant from a "chat tool" into "an intelligent agent that can execute tasks, build applications, and work continuously."

What is the essence of this move? It turns AI from a tool "you open when you need it" into an interface you live and work inside every day.

Second: Meta's Muse and Muse Charm.

In the same days, at its Connect conference, Meta announced Muse Charm — a small wearable device built around its AI assistant, Muse. The plan: bring Muse into smart glasses within the next few months, so a user wearing the glasses can talk directly to Muse about whatever they see.

(Media nicknamed it an "AI pet," but its real positioning is a wearable hardware gateway to the Muse agent.)

What is the essence of this move? It shifts AI's gateway from "screen" to "wearable" — no longer "look down at your phone to use AI," but "AI is right in front of your eyes."

Put the two together and it becomes clear:

Their shared ambition is the same: to make AI the thing you open first and keep open always. Like today's operating system — you don't "remember to open Windows"; you're always inside it.


4. Where the Two Fronts Meet

By now, each front is clear. But they actually converge at one point.

"Standards" teach people how to use AI; "gateways" contest the first AI people open.

These two seem unrelated but are causally linked:

So this is not a division of labor where "China does education, America does products." Both sides' moves ultimately point at the same goal: defining how the next generation defaults to living with AI.

China's path is "top-down": set standards first, then push into classrooms, using policy power to spread the literacy framework.

Global giants' path is "bottom-up": seize the gateway first, let habits form through use, and let habit gradually become de facto standard.

Which path wins? Too early to say. But one thing is certain: they will meet. When China's literacy standards meet the global giants' product gateways, the question "who defines AI literacy" becomes very concrete.


5. Three Questions for Readers

As an ordinary person, what do these grand competitions have to do with you? Quite a lot. Because they will ultimately determine what your children learn, use, and are judged by at school.

So when you next see "education AI" news, don't just look at "what new product came out." Ask three questions first:

Question one: Who is setting the standard? — Policy, a framework, or some product's usage habit? Whoever sets the standard gets the outcome leaning their way.

Question two: Who is holding the gateway? — A phone, a computer, or glasses and wearables? Whoever holds "the first AI you open" holds default authority.

Question three: Is my (or my child's) "literacy" sufficient? — Not "can you use the tool," but "can you judge, can you create, can you take responsibility." That is exactly what "know, use, discern, create, be responsible" answers.

Think through these three, and you can basically judge: is a piece of education-AI news "real progress," or "a passing trend"?


Closing

Back to that timetable.

From September 8 to September 28 — just three weeks — we saw moves across six dimensions at once: policy, standards, academia, industry, hardware, software. Such density is no coincidence; it is a signal.

It signals that education AI is moving from the stage of "whether it exists" into the stage of "who gets to decide."

And that "who gets to decide" ultimately answers not a technical question, but an older one: what kind of people do we want to cultivate?

Tools will keep updating; gateways will keep migrating. But if "literacy" is done solidly — if the next generation knows what AI is, can use it, can discern it, can create with it, and is willing to take responsibility for using it — then no matter how gateways change or standards are contested, the human remains the one making the call.

That is probably the answer every standard and every product ultimately cannot avoid.


References

  1. Xinhua: "Delivering Education That Satisfies the People — SCIO Press Conference on Building an Education Powerhouse During the 15th Five-Year Plan," 2026-09-23
  2. CNR: "2026 AI Large Model CEO Roundtable: AI Enters the Real World — Where Are Enterprises Stuck?" 2026-09-23
  3. Science and Technology Daily: "Domestic AI Education Agent Appears at UNESCO," 2026-09-10
  4. Sina Finance: "The Battle for Gateway, the Test of Monetization, the Game of Ecosystem: Three Layers of Speed in Online Education's AI Transformation," 2026-09-20
  5. Sina Finance / TechWeb: "TAL's Teacher AI Assistant 'Jiuzhang Teacher' Launches Over 170 Agents, More Than 300,000 Registered Users," 2026-09-24
  6. Hong Kong Education Bureau: Circular No. 107/2026, "Training for Schools on Implementing AI Education Plans," 2026
  7. ifeng Tech: "Meta Releases 'AI Pet' Device Muse Charm," 2026-09-24
  8. Sina Finance: "Microsoft Releases New Copilot, Introducing Home, Code and Autopilot," 2026-09-25
  9. UNESCO, Digital Learning Week 2026 official information, 2026-09