Tashkent, Uzbekistan (UzDaily.uz) — The World Media Summit in Beijing moved the AI debate from talk to rules. For newsrooms in Central Asia and the wider Global South, the question is whether we will help write them.
When nearly 500 media leaders from 112 countries met in Beijing for the 7th World Media Summit, the theme, "Media in the Intelligent Age: From Consensus to Action," signalled impatience. The industry has spent three years talking about artificial intelligence. Now it wants rules.
That is the right instinct. But rules help only if the people who must live by them take part in writing them, and only if they protect the one asset no algorithm can produce: trust.
The tools are no longer the question
Few in Beijing disputed that AI already touches almost every stage of news production: drafting, translation, editing, summarisation, recommendation and distribution. Xinhua showed an AI information agent with source tracing and verification functions, alongside two humanoid robot anchors. The technology is here, and it will keep getting cheaper.
The cost of error is what changes the conversation. Generative AI makes it easy to fabricate a voice, a video or a quote. One fabricated item that slips past an editor can undo credibility a newsroom spent decades building. Marek Traczyk of the Polish Media Association put the problem in one line: "Speed is not accuracy, everyone can broadcast, but who takes responsibility?" Kamel Mansari of Algeria's Le Jeune Indépendant went further, calling unconditional reliance on automated content "editorial suicide."
Reuters president Paul Bascobert made the same point from the other end of the industry: the tools have changed dramatically, but the principles have not. The most important element of journalism is still journalism done by humans.
The promise is real, and it should not be lost in the warnings. Used well, AI lets a small newsroom translate faster, summarise long official documents, check figures against databases and reach readers in more languages than its staff could cover alone. Xinhua president Fu Hua said media should make sure all media professionals have a fair chance to benefit from AI. AFP chairman Fabrice Fries made a similar point: the gains from this technological revolution should reach far beyond a handful of companies or countries.
Both statements are easy to agree with. The harder question is what it takes to make them true for a newsroom of twenty people in Tashkent, Dushanbe or Bishkek, rather than for a global agency with its own engineering department.
The gatekeepers have moved
A second uncomfortable truth came out of the panel on influence, AI and trust. Newsrooms no longer decide alone what audiences see. Platforms and their ranking systems decide, and so do creators with millions of followers. Vsevolod Pulya of RT China said plainly that big tech companies have become the new gatekeepers.
One video host on the panel admitted that content made with real money and effort often loses to "amusing false content." For an editor, that is the market failure in a sentence. Verification is expensive, falsehood is cheap, and the algorithm does not care which is which.
The panel could not agree on who should be accountable for misleading AI content: the creator, the publisher, the platform or the AI vendor. It concluded that responsibility cannot rest with any single gatekeeper. That may be true for the ecosystem as a whole. For a publisher, though, it is no reason to wait. Our name sits on what we publish. Whatever tools produced a text, the accountability is ours.
Why younger readers look elsewhere
The same panel asked why younger audiences increasingly follow individual creators rather than established news brands. The answers were less about technology than about tone. Mexican creator Diego Ruzzarin, who has more than 10 million followers across platforms, argued that lived experience and cultural connection matter more than an institutional logo. Pulya urged traditional media to engage with popular culture and said real emotional attachment is something every outlet should learn to build.
It would be easy to dismiss this as advice for entertainment channels. It is not. In Uzbekistan, as in much of the region, a large share of readers now get their news first from Telegram channels and social media accounts run by individuals. Many of them are fast and engaging; far fewer of them correct their mistakes in public. Established publishers cannot win that contest on speed alone, and we should not try to win it by lowering our standards.
What we can do is combine what creators do well with what we do well: a human voice, explanation instead of bare headlines, formats built for the phone, and the discipline of checking before publishing. AI can help with the first part, by turning a long report into a clear explainer or a short video script. Only editors can guarantee the second.
Who writes the standards matters
The summit's most concrete outcome was the Global Media Initiative for the Joint Development of AI Application Standards. Twenty-two founding participants, including Xinhua, TASS, Bernama and South Africa's Independent Media, plan to develop standards over roughly two years and test them in working newsrooms. The participants also backed a global database of AI applications for media and called for sharing open-source models, training resources and verification practices.
Delegates from Africa, Asia and small island states pressed the point that matters most for outlets like ours. Aminath Namza of the Maldives' Public Service Media asked how smaller nations can have a meaningful voice in shaping AI, noting that most models learn mainly from societies with a larger digital footprint. Viasen Soobramoney of South Africa's Independent Media argued that the Global South must take part in developing, governing and owning these technologies rather than simply consuming them. Wong Chun Wai of Bernama added that access to basic AI tools is not enough: newsrooms need to invest in advanced tools and train people to use them well.
From Tashkent, none of this is abstract. We publish in Russian, Uzbek in two scripts, and English. For most commercial models, Uzbek remains a lower-resource language. Machine translation of economic news still needs a human who knows the difference between a presidential decree and a draft resolution, or between a tax rate and a tax base. If the standards that come out of this process assume English-language newsrooms with large technology budgets, they will be standards for someone else.
The initiative is also led largely by state news agencies. Its standards will carry weight only if commercial, independent and regional publishers from many editorial traditions can join the drafting and testing on equal terms. A standard that most of the industry did not help shape is a recommendation, not a standard.
Cooperation that is practical, not ceremonial
Some of the most useful proposals in Beijing came from African delegates, and they were notable for being specific. Soobramoney proposed cooperation in five areas: AI-assisted translation, multimedia production, journalist training, content verification and the exchange of technical expertise. Agnes Kalekye Nguna of the Kenya Broadcasting Corporation called for dialogue among civilizations on an equal footing. Ghislain Ruffin Etoughet Nzuet of the Gabon Broadcasting Corporation said the summit gave African countries a platform to tell their own stories, noting how little the rest of the world knows about the continent.
Najeh Missaoui of Tunis Afrique Presse described the standards initiative as a move from talking about AI to joint action and rule-setting. Mohammed Alyami of the Union of OIC News Agencies said he hoped the standards would be usable by media organizations of every size and level of resources.
That last condition is the test the whole initiative will be judged by. Central Asian media know the feeling Etoughet Nzuet described: the world often learns about our region through other people's reporting, written in other languages and shaped by other priorities. AI could make that worse, if models trained on outside sources become the default way the world reads about us. Or it could make it better, if our own reporting in our own languages becomes part of the data and our own editors take part in setting the rules. Africa's list of five cooperation areas would serve Central Asian newsrooms just as well, among ourselves and with partners from every direction.
What smaller newsrooms should do now
We do not need to wait two years to act. Five steps are within reach of any newsroom today:
- Write our own AI rules first. Decide what may be automated, what must always be checked by a person, and how AI-assisted material is labelled. Publish those rules so readers can hold us to them.
- Treat verification as part of the product, not a cost. In an environment flooded with synthetic content, being checkable is a competitive advantage.
- Take part in standard-setting. National news agencies, publishers' associations and journalism schools in our region should apply to join the two-year process and offer their newsrooms as test sites.
- Cooperate regionally on language. Central Asian outlets share languages, markets and stories. Shared glossaries, translation memories and verification resources would cost each of us little and help all of us.
- Invest in people, not just licences. A subscription to an AI tool is not a strategy. Editors who understand both the tool and the story are.
Trust is the infrastructure
Abdul Hakim of Indonesia's Antara News Agency summed up the summit better than any communiqué: "Technology makes media faster. Intelligence makes media smarter. But trust makes media matter."
The stakes go beyond our industry. At a side forum on smart media, speakers linked responsible media and technology cooperation to global security, food and energy security, climate and sustainable development. Xinhua deputy editor-in-chief Ren Weidong said media should not be "bystanders or onlookers" when the world faces these challenges. In a crisis, a fabricated video or a mistranslated statement can do real damage within hours. Accurate, verified reporting is part of how societies keep their footing.
Abdusaid Kuchimov, director general of Uzbekistan's National News Agency, said in Beijing that information should build bridges between people, not walls. AI can help us build those bridges faster and in more languages. It cannot decide what is true, and it cannot take responsibility when we get it wrong. That part of the job is still ours, and it should stay that way.
