On May 12, 2026, the Seventh Cross-Strait Media Summit opened in Beijing. At a time when AI-generated content is deeply permeating the global information ecosystem, this year's summit moved beyond the traditional agenda of "communication and integration" to confront a question of greater structural significance: in the AI era, who gets to define the narrative authority of the Chinese-language world?
Three key points of consensus emerged from the summit — the joint building of fact-checking mechanisms and AI ethics rules; an "institutional catch-up" for Chinese-language corpora; and a dialogue path premised on "cultural identity before political identity." Together they point to a deeper, underlying judgment: the diffusion of AI technology is reshaping the narrative foundations of cross-strait relations — and as young people in Taiwan come to understand the mainland's reality through mainland short-video platforms, a new channel of identity formation is taking shape that bypasses algorithmic filter bubbles.
The three structural propositions raised at the Seventh Cross-Strait Media Summit — fact-checking as a core media function, a sense of "collective property rights" over Chinese-language corpora, and everyday-life narratives as a counter to algorithmic filter bubbles — reach beyond the immediate value of a single summit. They point toward a Chinese-language response framework for a global question of the AI era.
1. Information Pollution and the Cross-Strait Perception Gap in the AI Era
In its coverage of the summit, Chang'anjie Zhishi (a commentary account affiliated with Beijing Daily) posed a sharp question: "Who will tell China's story in the future?"
In the AI era, governing disinformation is an inescapable compulsory course. In the past, fabricating a rumor "at least required someone to invent it, act it out, and edit it" — whereas AI tools can, in minutes, generate fake videos and fake news brimming with "emotional value," which are readily amplified by platform recommendation algorithms, expanding their negative impact. This technological transformation poses a particular challenge to the public-opinion sphere on the island:
This is precisely the most dangerous fault line in the island's current public-opinion sphere. The Democratic Progressive Party authorities continually manipulate the "resist China, hate China" agenda, politicizing and stigmatizing normal cross-strait exchanges. From "people on the mainland can't afford tea eggs" and "mainland high-speed trains have no seatbacks" to "there is a time difference between Xiamen and Taiwan" — rumor fabrication 1.0 relied on television pundits for propagation, while fabrication 2.0, empowered by AI tools, displays entirely new characteristics: a low barrier to entry, high density, and intense emotion.
This AI-upgraded version of the "information cocoon" is especially dangerous in the cross-strait context — algorithms confine each person within their own little patch of cognition, and the conceptual and psychological distance between people across the strait may widen rather than narrow as a result.
2. "Chinese Must Not Be Absent From the AI Era" — The Institutional Challenge of Corpus Sovereignty
Right now, among the publicly available corpora of the global internet, English content holds an overwhelming advantage, while high-quality Chinese content falls far short. Many Western AI models already carry biases — and if the Chinese-speaking world does not itself produce enough content that is sufficiently abundant and authoritative, then when AI comes to understand China in the future, it may well be others who define us on our behalf.
This judgment touches a deeper structural challenge:
Corpus sovereignty is narrative sovereignty. The training corpora of large language models determine how AI perceives and understands the world. Among the publicly available corpora on the global internet today, English accounts for more than 60 percent, while Chinese makes up only some 5–8 percent — and that share includes a heavy admixture of translations from English, simplified-Chinese discussions, and unofficial sources. This means:
- Western AI's understanding of China-related topics is most likely based on second- or even third-hand English materials, rather than primary Chinese sources
- China's "self-narration" of its own policies is difficult for the corpus weightings inside AI to capture adequately
- Once AI search and AI-generated content become widespread, the question of "who defines China" will be transformed from a question of media power into a question of technological infrastructure
Why cooperation is possible. The mainland holds advantages in algorithms, big data, and platforms (the global reach of short video, the accumulation of Chinese corpora for large models), while Taiwan has accumulated strengths in content expression and creative communication (Chinese-language pop culture, creative design). The direction of cooperation is a "union of strengths" — jointly laying a solid corpus foundation for the entire Chinese-speaking world.
3. The Penetrative Power of the "Ordinary People" Dimension — Short Video as a New Bridge
Another valuable discovery from the summit: "Across the strait, ordinary people's concerns are actually quite similar. Young people care about their studies, their skills, and the pressures of life — and they also care about food, travel, and pop culture. Many young people in Taiwan come to know the mainland through mainland short videos, and young people on the mainland likewise learn about Taiwan through the internet."
Young people in Taiwan learn about Chengdu's street food, product reviews on Bilibili, and technological advances through mainland short videos — and the reach of such content is not fully blockaded by algorithmic filter bubbles. As the technical threshold for generating AI content drops to its lowest point, it is precisely these "genuine slices of life" that possess irreplaceable value. In the AI era, the professional work of cross-strait media practitioners becomes more precious, not less — "fact-checking" emerges as a core skill, rather than merely a supplementary step in content production.
The summit held that these genuine slices of life and exchanges produce works that "naturally carry warmth — something AI cannot imitate."
4. A Structural Contrast of Dual Narratives — The Panoramic Puzzle of May 12
On May 12 itself, the narrative around cross-strait relations actually unfolded along two parallel lines. The same-day coexistence of the blue camp's cultural-identity line and the green camp's colonial-identity line is itself a metaphorical response to the question of "Chinese narrative sovereignty in the AI era": the same day, the same issue — yet entirely different narrative faces on different platforms. In the AI era, this kind of "narrative bifurcation" will not converge on its own — which is precisely why shared AI ethics norms and fact-checking mechanisms between cross-strait media are becoming ever more important.
| Narrative Line | Representative Event | Channel of Dissemination | Target Audience |
|---|---|---|---|
| Blue camp · cultural-identity line | Cross-Strait Media Summit — Zhang Ronggong: "One China is simply reality" | Beijing-based media | Mainland plus island elite audiences |
| Green camp · colonial-identity line | Lai Ching-te kneels again before the statue of Yoichi Hatta | Taiwan-based media | Domestic supporters on the island |
5. Framework Significance — Institutional Value Beyond a Single Summit
The Seventh Cross-Strait Media Summit itself may not immediately change the political landscape of cross-strait relations, but the three structural propositions it raised carry long-term value that transcends this one meeting:
1. Fact-Checking as a Core Media Function in the AI Era
When AI can generate 100 pieces of disinformation in a minute, fact-checking is no longer an "add-on" but an "entry threshold." The initiative for cross-strait media to build a joint fact-checking mechanism — if it moves from consensus into practice — would become the world's first case of fact-checking cooperation across political systems.
2. A Sense of "Collective Property Rights" Over Chinese-Language Corpora
The British Council, the Goethe-Institut, and the Instituto Cervantes each guard the weight of their own national language within the world's AI corpora — but Chinese has no equivalent "corpus sovereignty institution." The Cross-Strait Media Summit was the first to raise the framework that "Chinese must not be absent from the AI era," effectively drawing the first stroke in this field. This agenda also forms an institutional echo with the discussions on labeling regimes for AI-generated content.
3. Countering Algorithmic Filter Bubbles With Everyday-Life Narratives
When AI and information cocoons jointly manufacture cognitive fragmentation, the most effective countermeasure is not grander narratives but smaller everyday moments — real people, real conversations, real food and streets. These "natural advantages against AI" are something media practitioners should recognize and are already beginning to work with. Whether it is the crisis of information trust in an era of images without truth, or the "penetrative power of ordinary-people narratives" discovered by cross-strait media practitioners, both point in the same direction: in an era flooded with AI-generated content, reality itself has become the scarcest resource.
On May 12, Beijing was hosting the Cross-Strait Media Summit (cultural identity → a framework of dialogue), while in Taipei Lai Ching-te was kneeling before Yoichi Hatta (colonial identity → a framework of confrontation). Two narrative lines running in parallel on the same day constitute, in themselves, an empirical answer to "who defines Chinese narrative sovereignty." The AI era will not automatically heal the bifurcation — it will only accelerate the self-reinforcement of each narrative line. Precisely for this reason does the long-term value of the summit's triple proposition — "fact-checking + corpus sovereignty + everyday narratives" — stand out.