Across Asia, public interest media are turning to artificial intelligence (AI) to optimize under-resourced newsrooms. However, these outlets face a high-stakes reality: publishing AI-generated errors, factual inaccuracies, or “AI slop” risks undermining the public trust and credibility central to their legitimacy. At the same time, these media are also aware that they are operating in authoritarian contexts where AI-induced mistakes can lead to legal and political persecution.

Hence, independent outlets in the Asia-Pacific – the media that produce journalism essential to democratic accountability – exercise “principled caution” when deploying AI. They exercise this caution not just because of a desire to maintain their reputation for accuracy. They also do so because there is a real risk of forced state closure, the freezing of assets, or the imprisonment of staff. 

This is the core finding of Asia Centre’s latest report, “Principled caution: How AI is used in public-interest media in Asia,” published in partnership with DW Akademie and supported by the German Federal Ministry for Economic Cooperation and Development.

Evaluating media ecosystems across eight Asian countries – Afghanistan, Bangladesh, Bhutan, Cambodia, Indonesia, Myanmar, Pakistan, and Uzbekistan – the study reveals how journalists are managing AI adoption. Rather than blind adoption, Asia’s independent public interest media are approaching AI through a careful and highly selective strategy. In structurally precarious environments where reporting errors can invite censorship, legal action, or operational shutdown, outlets consistently prioritize human editorial oversight over full-scale automation.  

The report identifies several key findings:

First, AI usage in newsrooms remains assistive rather than transformative. Outlets primarily use generative AI tools at the preliminary and final stages of editorial workflows, for tasks such as ideation, document summarization, social media adaptation, and headline polishing. The central stages of journalistic work – including original field reporting, investigative narrative drafting, and editorial judgment – remain strictly human-driven.

Second, severe language constraints mean that major global Large Language Models underperform in low-resource Asian languages, including Bahasa Indonesia, Bangla, Burmese, Chittagonian, Dzongkha, Farsi, Jingpo, Khmer, Pashto, and Urdu. These platforms frequently mistranslate context-sensitive terms, flatten cultural nuances, and generate wrong outputs. While translation is the most widely adopted AI task, outputs require labor-intensive human correction.

Third, exiled media from countries like Afghanistan, Bangladesh, Myanmar, and Pakistan face distinct operational threats including digital surveillance, internet blackouts, cross-border security risks, and severe financial instability. These pressures leave little mental or institutional bandwidth to experiment with new technologies. Yet, securely tailored automated tools for transcription, translation, and media-monitoring offer an unrealized opportunity to engage diaspora networks and report on repressed homelands without compromising journalists’ digital safety.

Fourth, a sharp institutional deficit persists across newsrooms. Over 60 percent of surveyed newsrooms report that AI adoption occurs via ad-hoc, individual staff experimentation rather than cohesive organizational strategies. Structural gaps, such as the absence of formal editorial guidelines, missing data protection protocols, and a heavy reliance on Big Tech platforms, leave independent media vulnerable to algorithmic bias, privacy violations, copyright risks, and state surveillance.

Fifth, restrictive political and legal environments across Asia impede responsible AI adoption. While state authorities deploy automated systems for online monitoring, surveillance, and synthetic media campaigns, regulatory ambiguity is routinely weaponized to clamp down on media freedoms. Public interest newsrooms face a stark structural asymmetry; political actors deploy AI at scale to amplify state narratives and disinformation, while underfunded independent outlets lack affordable verification tools to counter these attacks effectively. 

Looking ahead, AI integration in Asian newsrooms cannot rely solely on off-the-shelf tools built by Western tech monopolies. Long-term media sustainability requires localized AI infrastructure, enforceable newsroom governance frameworks, and open-source language models tailored to the region’s complex political and linguistic contexts. Without this shift toward technological sovereignty, regional media risk becoming mere data points for foreign algorithmic models that operate without local democratic stakes. We must move beyond being passive consumers of black-box solutions and start architecting a digital future that respects our region’s unique socio-political boundaries.

To achieve this, international donors and media development actors must pivot away from short-term, surface-level training sessions. Instead, capital must be redirected toward regional and local entities to lead core technological infrastructure, secure data storage systems, and create collaborative AI hubs specifically designed to serve independent and exiled media. This requires reimagining “capacity building” as the construction of resilient, shared digital architectures that treat cybersecurity as a foundational reporting tool. By financing regional and local based collective defense, we can turn isolated public interest newsrooms into fortified digital outposts capable of weathering persistent state-sponsored surveillance.

Ultimately, the future of journalism in Asia will not be determined by the speed of algorithmic tools, but by whether newsrooms can harness AI without compromising the trust, accuracy, and independence that underpin public interest reporting. In an era of synthetic media floods, human-centric rigor is not a hurdle to innovation – it is our ultimate competitive advantage. Those who prioritize ethical safeguards now will be the only ones to retain audience loyalty when the inevitable global backlash against AI-generated misinformation arrives.

This op-ed is based on Asia Centre’s new report, “Principled caution: How AI is used in public-interest media in Asia”. Download the full report here. For more information about the Asia Centre, visit asiacentre.org.