It’s a busy afternoon, and the journalists inside are chasing three stories simultaneously — a government budget controversy, a community health crisis in a densely populated neighborhood, and a breaking story about flooding in a rural state that most international outlets won’t touch. They’re working in a mix of English and Yoruba, their sources speak in pidgin, their audience consumes news on low-bandwidth mobile connections, and the cultural context of every story they tell is shaped by decades of post-colonial history, local political dynamics, and community relationships that no outsider could fully grasp from a distance.
Now imagine that newsroom being handed an AI journalism tool built in San Francisco, trained primarily on English-language Western news content, optimized for broadband internet connections, and calibrated to the editorial standards and audience expectations of American or European media organizations. How well does that tool serve the journalists in that Lagos newsroom? How accurately does it understand the linguistic nuances of their reporting? How well does it contextualize the stories they’re telling? How relevant are its content suggestions to the communities they serve?
The answer, if we’re being completely honest, is: not very well. And that gap — between what Western AI tools are built to do and what news organisations in developing countries actually need — is at the heart of one of the most important debates in global journalism right now.
Should news organisations in developing countries simply adopt the AI tools that Western tech companies are building and offering to the world? Or should they invest the considerable time, resources, and effort required to build their own? It’s a question that goes far deeper than technology. It touches on sovereignty, identity, economic equity, and the fundamental question of who gets to shape the information ecosystems of the world’s most rapidly growing populations.
The AI Tool Landscape That Exists Today
Before we can evaluate whether developing country news organisations should build their own AI tools, we need to understand the landscape of tools that currently exists and what it was actually built for. The AI tools dominating the journalism technology conversation right now — tools for automated content generation, transcript processing, audience analytics, misinformation detection, and news aggregation — have been developed overwhelmingly by technology companies based in the United States and Western Europe.
OpenAI, Google, Microsoft, Meta, and a constellation of smaller AI startups have produced tools that are powerful, increasingly accessible, and in some cases genuinely useful for journalists anywhere in the world. These tools were trained on datasets that are, by any measure, enormously skewed toward English-language content and Western cultural contexts. The Common Crawl dataset that underlies many large language models contains predominantly English content. The news datasets used to fine-tune journalism AI tools draw heavily from American and European publications. The editorial standards, story structures, and audience expectations baked into these systems reflect the journalism cultures of the Global North far more than those of the Global South.
This is not a conspiracy or an act of deliberate exclusion. It’s a consequence of where AI development has been concentrated, which reflects where capital, technical talent, and research infrastructure have historically been concentrated. But the consequences for news organisations in developing countries are real and significant, regardless of the intent behind them.
Language Is the Most Obvious Problem
Let’s start with the most immediately visible challenge: language. The world is a staggeringly multilingual place. There are over 7,000 languages spoken globally, and the majority of the world’s people conduct their daily lives in languages that are dramatically underrepresented in AI training data. Africa alone has over 2,000 distinct languages and dialects. India has 22 officially recognized languages and hundreds more in common use. Southeast Asia, Latin America, and the Pacific are similarly linguistically diverse.
News organisations in developing countries often operate in these underrepresented languages. A newspaper in Ethiopia might publish in Amharic. A radio station in Bangladesh operates in Bengali. A digital news startup in Mexico’s Oaxaca state may publish in Zapotec indigenous languages alongside Spanish. For these organisations, AI tools trained primarily on English-language data are not just suboptimal — they’re often genuinely unusable for core journalistic functions.
Automated transcription tools that work beautifully for English-language interviews produce inaccurate or nonsensical output for Swahili or Tagalog. Content moderation tools trained on English social media may fail to identify misinformation or harmful content expressed in regional dialects. Translation tools that handle major world languages adequately may completely fail for less widely spoken languages, producing outputs that range from inaccurate to embarrassing to genuinely harmful when applied to sensitive journalistic contexts.
The language problem is not just about linguistics. Language carries culture, history, and meaning in ways that are inseparable from the words themselves. A news organisation reporting in Hausa is not just reporting in a different language from English — it’s operating within a completely different cultural framework, storytelling tradition, and audience relationship. No AI tool trained on Western news content can understand that framework without being specifically trained to do so.
The Cultural Context Gap Is Deeper Than Most People Realize
Language is the visible tip of a much deeper iceberg. Beneath it lies an equally significant cultural context gap that affects how AI tools understand, generate, and evaluate journalism from non-Western contexts. Journalism, at its core, is a culturally situated practice. The stories that matter, the sources that carry authority, the narrative structures that resonate with audiences, and the ethical frameworks that guide editorial decisions are all shaped by specific cultural, historical, and political contexts.
Western AI tools carry within them deeply embedded assumptions about what journalism looks like. They tend to assume a particular relationship between journalists and government — one shaped by Western liberal democratic traditions in which the press is explicitly adversarial to state power. They assume certain story structures — the inverted pyramid, the objective third-person voice, the quote-heavy news story format that dominates American and British journalism. They assume certain ideas about what constitutes a reliable source and what kind of evidence merits inclusion in a news story.
None of these assumptions are universal. In many developing countries, journalism has evolved within different political contexts, different cultural storytelling traditions, and different audience relationships. Community radio in West Africa often operates within a fundamentally different relationship between journalist and community than the adversarial model that dominates Western journalism culture. Storytelling traditions in many Asian and African journalism cultures use narrative structures that differ significantly from the inverted pyramid. The authority attributed to different types of sources — elders, community leaders, oral testimony, religious figures — varies enormously across cultural contexts.
When an AI tool that doesn’t understand these differences is used to evaluate, generate, or assist with journalism in these contexts, it produces outputs that are at best culturally tone-deaf and at worst actively distorting of the journalism it’s supposed to support.
Data Sovereignty and Who Controls the Information
There’s a dimension of this conversation that goes beyond technical performance and touches on something more fundamental: data sovereignty. When a news organisation in Kenya, Indonesia, or Peru adopts a Western AI journalism tool, it typically has to feed data into that tool — audience data, content data, behavioral data, editorial data. Where does that data go? Who owns it? How is it used? What happens to it if the company providing the tool changes its terms of service, gets acquired, or simply decides to discontinue the product?
These are not paranoid hypotheticals. The history of technology adoption in developing countries is full of examples where initial access to a tool or platform came with terms that were, in the long run, exploitative or constraining. Social media platforms that initially gave developing country publishers free reach and then changed their algorithms to demand payment for the same reach. Technology companies that harvested data from emerging market users to improve products that those users would ultimately pay more to access. Platform dependency that left organisations vulnerable when terms changed or services were discontinued.
Data sovereignty — the principle that communities and nations have the right to control the data generated within them and about them — is increasingly recognized as a critical issue in global technology governance. For news organisations, which are custodians of enormous amounts of sensitive information about sources, communities, and political actors, the question of where their data goes when they use a Western AI tool is not just a privacy question. It’s a question of journalistic integrity and source protection.
The Economic Dependency Trap
There’s also a straightforward economic argument for why developing country news organisations should be cautious about wholesale adoption of Western AI tools. When you build your journalism infrastructure around tools provided by external technology companies, you create a dependency that can be economically costly and professionally constraining.
Subscription fees for AI journalism tools — even at discounted rates sometimes offered to developing country organisations — represent a significant ongoing cost that flows out of local journalism ecosystems into the coffers of Silicon Valley or European technology companies. As these tools become more integral to newsroom operations, the switching costs rise, making organisations increasingly locked in to whatever pricing and terms the provider decides to impose.
There’s also the innovation dependency problem. When a news organisation’s technology roadmap is entirely determined by what Western tool providers choose to build and prioritize, it loses the ability to shape its own technological development in directions that serve its specific needs. The features that matter most to a newsroom in Nairobi or Hanoi are not necessarily the features that matter most to a newsroom in New York or London. If the technology is built elsewhere, the priorities of those distant contexts will always dominate the development roadmap.
What Building Your Own Actually Means
When we talk about developing country news organisations building their own AI tools, it’s important to be realistic about what that means and what it requires. We’re not necessarily talking about every individual news organisation building a complete AI system from scratch. That would be neither realistic nor efficient. What we’re really talking about is a spectrum of possibilities that range from adaptation to full development.
At one end of the spectrum, news organisations can adapt existing open-source AI models to better serve their specific needs. Training an open-source language model on locally relevant data — news content in regional languages, culturally specific training examples, locally calibrated editorial standards — is far more achievable than building a model from scratch and produces tools that are far more contextually appropriate than unmodified Western tools.
In the middle of the spectrum are collaborative development models — journalism organisations, academic institutions, and technology partners in a region pooling resources to build shared tools that serve their common needs. Several such initiatives are already underway. African language AI research consortia, regional journalism technology networks, and development-focused AI research institutes are all working on tools that address the specific needs of journalism in their regions.
At the other end of the spectrum — and this is realistic for only the largest and best-resourced news organisations in developing countries — is genuine ground-up AI development. A handful of major news organisations in countries like India, Brazil, and South Africa have the resources and technical talent to develop proprietary AI tools, and some are doing exactly that.
The Case of India: A Middle Path That Works
India offers one of the most instructive examples of how a developing country’s media ecosystem can navigate AI adoption with genuine sophistication. India’s news media landscape is extraordinarily diverse — it operates in 22 official languages, serves audiences ranging from rural agricultural communities to hyper-sophisticated urban digital natives, and encompasses both ancient journalistic traditions and cutting-edge digital innovation.
Several Indian news organisations have taken a notably pragmatic approach to AI tools. Rather than simply adopting Western tools wholesale or attempting to build entirely from scratch, they’ve pursued adaptation strategies — taking open-source models and fine-tuning them on Indian-language news content, working with Indian AI research institutions to develop tools calibrated to Indian journalistic contexts, and advocating within international technology forums for AI development that better serves the Global South.
The Wire, one of India’s leading independent digital news platforms, has been thoughtful about AI integration in ways that preserve its editorial independence and cultural specificity. Similarly, regional language news organisations in states like Tamil Nadu and Bengal have been developing AI tools for translation and transcription that are specifically calibrated to their linguistic and cultural contexts. India’s relatively strong technology sector and large pool of AI research talent make it better positioned than many developing countries to pursue this kind of sophisticated adaptation — but the approach offers models that can be adapted for less technically endowed contexts as well.
Africa’s Unique AI Challenge and Opportunity
Africa presents both the most acute version of the AI adoption challenge and some of the most exciting examples of grassroots response. The continent’s extraordinary linguistic diversity — those 2,000-plus languages — means that the gap between Western AI tools and African news organisations’ needs is probably wider than anywhere else in the world. At the same time, Africa’s rapidly growing population of young, technically skilled people and its history of leapfrogging older technologies in favor of newer ones creates genuine opportunity for African-led AI development in journalism.
Initiatives like Masakhane, a grassroots organization developing natural language processing tools for African languages, represent exactly the kind of locally-led AI development that could transform African journalism’s relationship with technology. Masakhane has worked to develop machine translation, speech recognition, and text generation capabilities for dozens of African languages — capabilities that Western AI companies have largely ignored because the commercial incentives to serve African-language markets are insufficient by their metrics.
Several African journalism organizations and press freedom groups have begun developing content moderation and misinformation detection tools specifically calibrated to African information environments — environments where the vectors of misinformation, the networks through which it spreads, and the cultural contexts that make certain false narratives persuasive are completely different from those in Western contexts. WhatsApp-based misinformation, which is a massive problem in African news ecosystems, requires detection and response strategies that are entirely different from those developed for Twitter-based misinformation in the United States.
Latin America’s Journalism Innovation Labs
Latin America offers another rich set of examples of developing region news organisations taking active roles in shaping their AI futures rather than simply receiving technology from elsewhere. The region has a vibrant ecosystem of journalism innovation labs — organizations like Chicas Poderosas, SembraMedia, and ICFJ’s network of Latin American partners — that are building locally relevant digital journalism tools and methodologies.
Brazil, with its large technology sector and sophisticated digital journalism ecosystem, has seen particularly interesting AI development for journalism. Several Brazilian news organizations have developed AI-powered fact-checking tools specifically designed for the Brazilian political and cultural context — tools that understand the specific claims, actors, and information ecosystems of Brazilian public discourse in ways that no American-built tool could without extensive local adaptation.
The broader lesson from Latin America is that regional collaboration — sharing resources, expertise, and development costs across organizations in multiple countries — can make AI tool development feasible for journalism ecosystems that couldn’t support it individually. A fact-checking AI tool developed collaboratively by organizations in Brazil, Argentina, Colombia, and Mexico can serve the entire region’s Portuguese- and Spanish-language journalism ecosystems in ways that are far more culturally and contextually appropriate than any imported Western tool.
The Open Source Path Forward
One of the most promising pathways for developing country news organisations to develop contextually appropriate AI tools without bearing prohibitive development costs is the open-source ecosystem. The rapid proliferation of high-quality open-source AI models — Meta’s LLaMA series, Mistral, and dozens of other open-source language models — has fundamentally changed the economics of AI development.
A news organisation in Southeast Asia no longer needs to build a large language model from scratch to have a powerful AI foundation to work from. It can start with an open-source model and invest the far more manageable resources required to fine-tune it on locally relevant data. This makes genuinely capable, locally adapted AI tools achievable for organisations with modest budgets and small technical teams.
The key investment required is not computational infrastructure but data — specifically, high-quality training data in relevant languages and cultural contexts. Building datasets of accurately annotated local-language news content, building collections of locally relevant editorial examples, and developing evaluation benchmarks calibrated to local journalistic standards are all achievable investments for organisations that commit to them. And the journalism organisations that invest in building these datasets will not just be improving their own tools — they’ll be making contributions to the broader open-source ecosystem that benefit the entire regional journalism community.
The Misinformation Dimension
The stakes of AI tool appropriateness become particularly acute when we consider the misinformation challenge. Misinformation is a global problem, but its specific characteristics vary enormously by context. The types of false narratives that spread most virally, the cultural and political dynamics that make them persuasive, the networks and platforms through which they circulate, and the effective strategies for countering them are all highly context-specific.
Western AI-powered fact-checking and content moderation tools are calibrated to Western misinformation ecosystems. They are built to detect the kinds of false claims, manipulated media, and coordinated inauthentic behavior that characterize misinformation campaigns in American and European political contexts. They are far less effective at detecting the kinds of misinformation that spread in developing country contexts — religiously motivated false narratives in South Asian social media, politically motivated ethnic incitement in African WhatsApp networks, health misinformation calibrated to local medical practices and folk beliefs.
For news organisations in developing countries that are trying to serve as bulwarks against misinformation in their communities, using AI misinformation detection tools that are miscalibrated to their specific context is not just suboptimal — it can be actively harmful. Tools that produce high rates of false positives in local languages, that miss culturally specific misinformation signals, or that flag legitimate local journalism as problematic based on criteria derived from Western editorial norms can undermine rather than support accurate journalism.
Building Capacity: The Human Side of the Equation
No discussion of whether developing country news organisations should build their own AI tools would be complete without addressing the human capacity question. Building, maintaining, and iterating on AI tools requires technical talent — data scientists, machine learning engineers, software developers — that is in short supply in the media sectors of many developing countries.
This is a genuine constraint, and it’s one that has to be addressed honestly. The gap between what organisations want to do and what they have the human capacity to do is real. But it’s also not insurmountable, and several approaches are making it more manageable. University partnerships between journalism schools and computer science or data science programs are creating pathways for journalism-oriented AI development that draws on broader technical talent pools.
Diaspora networks — technically skilled people from developing countries working in major technology centers who maintain connections to their home countries — represent an underutilized resource for technical capacity building. And regional journalism technology networks that allow organizations to share technical talent and development resources make sophisticated AI development achievable at scales that individual organizations couldn’t sustain.
The Role of International Development Funding
International development funding for journalism — from foundations, development banks, bilateral aid programs, and philanthropic organizations — has a crucial role to play in supporting locally-led AI tool development for news organisations in developing countries. This support needs to shift from primarily funding content production and business model development to also funding technology development and adaptation.
Historically, journalism development funding has been slow to recognize technology development as a core component of journalism capacity building. This is changing, but not fast enough. Organizations like the International Center for Journalists, the Global Forum for Media Development, and major journalism foundations are increasingly recognizing that the technological infrastructure of journalism in the Global South needs investment alongside the editorial and business dimensions. Funding specifically directed toward locally-led AI tool development — including support for training data collection, open-source model adaptation, and regional technology collaboration — would make an enormous difference for the journalism organizations trying to pursue this path.
The Political Economy of AI Adoption
There’s a political economy dimension to this conversation that rarely gets the attention it deserves. The choices that news organisations in developing countries make about AI tools are not made in a vacuum — they’re shaped by power relationships, institutional pressures, and incentive structures that often push toward adoption of Western tools even when locally developed alternatives might serve better.
When a major foundation funds a journalism innovation program and specifies that grantees should use a particular Western AI tool that the foundation has partnered with, it creates incentive structures that are difficult for cash-strapped local organizations to resist. When technology companies offer their tools for free or at dramatically subsidized rates to developing country news organizations, the short-term economic logic of adoption is compelling even when the long-term dependency costs are significant. When international journalism training programs are built around Western AI tools, they transmit not just technical skills but cultural assumptions about how journalism should be practiced.
Being aware of these political economy dynamics doesn’t mean rejecting international partnerships or declining available resources. It means approaching them with clear eyes, negotiating terms that preserve local control and data sovereignty, and building toward technological independence as a long-term goal even while making pragmatic short-term choices.
A Hybrid Strategy Makes the Most Sense
Having considered the arguments on all sides of this question honestly, the most defensible conclusion is not a binary one. The choice between “build your own” and “adopt Western tools” is a false dichotomy that serves neither the interests of developing country journalism organizations nor the interests of the global information ecosystem they’re part of.
The most sensible approach is a sophisticated hybrid strategy that uses Western tools where they serve well and work to develop locally appropriate alternatives where they don’t. For news organizations in developing countries, this means conducting honest assessments of which AI tools genuinely serve their specific linguistic, cultural, and journalistic contexts and which ones are being adopted primarily out of convenience, funding incentives, or FOMO about technological modernity.
It means investing in data collection and model adaptation even when full tool development isn’t feasible. It means participating in regional collaboration networks that pool resources for technology development. It means advocating loudly and persistently within international AI development forums for training data, development priorities, and investment that better serves the Global South. And it means building the internal technical capacity — even if slowly and modestly — that will make greater technological independence achievable over time.
Conclusion
The question of whether news organisations in developing countries should build their own AI tools rather than adopting Western ones ultimately resolves into a question about power, identity, and the future of journalism as a globally diverse practice. If we believe that journalism serves communities best when it is deeply embedded in those communities’ languages, cultures, and realities, then we must also believe that the technological tools that support journalism should reflect that same embeddedness. Western AI tools, however sophisticated and well-intentioned, cannot provide that without significant local adaptation — and adaptation that is led and controlled by the communities being served, not by the technology providers doing the serving.
Building locally appropriate AI tools for journalism in developing countries is not just a technical challenge. It’s an act of intellectual sovereignty and cultural self-determination. The news organisations that pursue it — strategically, collaboratively, and with clear eyes about both the difficulties and the stakes — are not just investing in their own operational efficiency. They are investing in the capacity of their communities to understand themselves accurately, to hold power accountable in culturally appropriate ways, and to participate in global information flows on their own terms rather than on terms imposed from elsewhere. That investment is worth making. The journalism of the global majority deserves nothing less.
Frequently Asked Questions
What are the biggest specific limitations of Western AI journalism tools for news organisations in developing countries?
The most significant limitations operate at several levels simultaneously. Linguistically, most Western AI tools perform poorly or not at all in the hundreds of languages spoken across Africa, Asia, and Latin America that are underrepresented in training data. Culturally, these tools embed assumptions about story structure, source authority, and editorial standards derived from Western journalism traditions that don’t translate well to different journalistic cultures. Technically, many tools are optimized for high-bandwidth connections and device capabilities that aren’t representative of audiences in developing countries. And in terms of data governance, using Western tools often means surrendering control over sensitive journalistic data to companies operating under foreign legal frameworks with little accountability to local communities.
Is it realistically possible for news organisations in developing countries to build their own AI tools given resource constraints?
Full ground-up AI development is only realistic for the largest, best-resourced news organisations. But the most important and achievable goal for most organisations is not building from scratch but adapting existing open-source AI models to local contexts — a far more achievable undertaking. This requires investment in local-language training data, technical partnerships with regional universities and research institutions, and participation in regional journalism technology collaboration networks. The open-source AI ecosystem has dramatically lowered the barriers to this kind of adaptation, making locally appropriate AI tools achievable for organisations that would never have the resources to develop full AI systems independently.
How can international funders and development organizations better support locally-led AI development for journalism in developing countries?
International funders need to shift from primarily funding content production and business model development to also funding the technological infrastructure of journalism in the Global South. Specifically, this means grants and programs specifically directed toward local-language training data collection, open-source model adaptation projects, regional journalism technology collaboration networks, and technical capacity building within journalism organizations. Funders should also be cautious about programs that effectively mandate adoption of specific Western tools, and should instead prioritize approaches that build local technical capacity and ownership. Longer grant timelines — recognizing that technology development requires sustained investment rather than short-term project funding — are also essential.
What role can regional collaboration play in making AI tool development more feasible for developing country news organisations?
Regional collaboration is probably the single most powerful mechanism for making locally appropriate AI development feasible for news organisations that couldn’t achieve it independently. By pooling data, sharing development costs, and building tools that serve the common needs of journalism ecosystems across multiple countries in a region, collaborative initiatives can achieve scale and capability that no individual organization could reach. Existing models like the Masakhane initiative for African language AI, regional journalism innovation labs in Latin America, and developing Southeast Asian journalism technology networks all demonstrate what collaborative approaches can achieve. The greatest need is for sustained funding and institutional support to build on these models and scale them to meet the full scope of the need.
How should news organisations in developing countries evaluate whether to adopt a Western AI tool or invest in local alternatives?
The evaluation should consider several key dimensions. First, performance: does the tool actually work well for the specific languages, formats, and contexts the organisation operates in, or is its performance significantly degraded? Second, data governance: what happens to the organisation’s data when it uses the tool, who owns it, and what protections exist for source confidentiality? Third, dependency risk: how locked-in does adoption make the organisation, and what are the costs of switching if terms change or the tool is discontinued? Fourth, alternatives: is there a locally developed or open-source alternative that could be adapted to serve the organisation’s needs with a manageable investment? And fifth, advocacy potential: does adopting the tool come with opportunities to advocate for the features and improvements that would make it genuinely serve the organisation’s community — or does the power relationship make such advocacy futile?
Kathy Jane is a writer who specializes in public administration and media communication. She has 17 years of experience covering these fields and keeping up with their main trends. Kathy holds a BSc and an MSc in Mass Communication, giving her the skills to explain government and media topics in clear, easy-to-understand language.
Leave a Reply