As AI sharpens media’s class divide, lessons from the global majority

“I hope the examples are not all about turning articles into bullet-point summaries.”

I said that half-jokingly to a colleague as we sat down to watch our umpteenth AI-in-journalism panel at a conference earlier this year. We were bummed as the panelists, who worked for some of the best-known newsrooms in the world, went on to discuss their adventures in bullet-point summaries.

It was a disappointment, but hardly a surprise because complex AI projects are hard to implement. The prompting is easy, but building data pipelines, workflows, and risk mitigation systems is endlessly hard. It’s difficult enough for bigger newsrooms; many smaller outlets just don’t have the resources to do it.

That’s the broken promise of AI. Instead of democratizing intelligence, the full potential of the technology is only available to the rich — those who can pay for an army of engineers and data analysts, not to mention for hardware and cloud infrastructure — sharpening the media industry’s already deep class divide. The rest are left to use AI for more mundane tasks like transcription, SEO headlines, and yes, bullet-point summaries. Valuable time-savers, but hardly revolutionary use cases.

No lifelines are coming from big tech. Not technologically, when most AI labs are releasing new models meant to top benchmark leaderboards, with agentic functions that may seem impressive on the surface, but are more opaque, harder to audit, and thus more difficult to use for journalistic purposes.

And certainly not financially. Big tech has only ever been interested in placating the big newsrooms with a pittance for licensing deals, and withevidence steadily building to show that audiences are not all that interested in using chatbots to find news, one could imagine that their already small appetite to do these deals would be shrinking further. Support from big tech for smaller newsrooms usually comes from one-time grants or training programs, often as part of project-based journalism development initiatives.

So what do you do in a world where the technology never centers you, the platforms never pay you, and there is no guarantee of visibility?

Many newsrooms in the Global Majority have been living in this world for years. This is in addition to the pressures some of them may face economically and politically in their local contexts. The ways these organizations have navigated this AI moment offer lessons for the rest of the world.

For example, I’ve worked with large news outlets with tiny data teams — think one to four people — such as ABS-CBN and Inquirer.net in the Philippines and Sin Chew Daily in Malaysia, who used AI to find deeper insights from their content and audience data. It’s similar to the excellent work being done by the Daily Maverick in South Africa, where they use AI to boost engagement and reader revenue. Even for newsrooms with small audience teams, using AI to surface insights from content and audience data can be an easy win because it doesn’t require complex infrastructure to operationalize.

News organizations in the Global Majority have had to contend with the relative lack of support for their languages from AI models, but some have been able to turn that into advantages. Hukum Online, a legal research platform in Indonesia, launched an AI search tool for its comprehensive Bahasa Indonesia knowledge base that has become a premium feature for its B2B subscribers. India’s Mathrubhumi, which publishes in the Malayalam language, is working on an ambitious project to turn its century-old archive into a dataset that can be used to train models to support the language and the script.

Using AI to extract value out of content isn’t limited to non-English languages. India’s Scroll has created a tool to scale their video publishing operations, while the head of their AI lab Sannuta Raghu has done pioneering work to create a framework for how newsrooms could implement liquid content. Speaking of which, I also created a prototype for how longform text could be turned into illustrated video using AI.

Newsrooms don’t have to go at it alone. In Latin America, El Centro Latinoamericano de Investigación Periodística (CLIP) manages a data commons platform that has since been enhanced by AI, allowing journalists across 21 countries to quickly search and cross-reference information.

Public datasets are especially valuable given how easy it is now to investigate them with the help of chatbots. Amid furor in the Philippines over flood infrastructure corruption, I used AI to vibe-code a script to scrape data from the government’s official website and posted it publicly as a spreadsheet. Media outlets and think tanks not only created their own investigations, but regular citizens loaded the data into a chatbot and to answer their own questions.

Because their conditions are structurally different, elite newsrooms need to stop being the model for navigating the AI moment for news organizations on the wrong side of the media’s class divide, which will only worsen given the direction of AI development.

Instead, we should look at the myriad examples from the Global Majority into how to find accessible wins that avoid heavy systems work, making them viable for organizations with limited resources. Value extraction from existing content and data, not expensive AI infrastructure, is where the most practical opportunity lies for most newsrooms. There also needs to be more experiments with data commons and shared technical infrastructure.

As more newsrooms become marginalized, perhaps it’s time to look to organizations that have long been working at the margins as operating models in a more constrained future.