Information Integrity Is Development Infrastructure

The UNU Global AI Network Insight Blog Series is launched to celebrate the Network’s two-year anniversary. In this series, members of the UNU Global AI Network share diverse insights at the intersection of artificial intelligence and sustainable development. Bringing together perspectives across regions and disciplines, the series explores both the opportunities and challenges of shaping AI for all. 

The views expressed in published articles of this series are solely those of the author(s) and do not necessarily reflect the positions of the United Nations University, UNU Macau, the United Nations, or affiliated entities. Publication does not imply endorsement or institutional approval. 

Authors are responsible for the accuracy and originality of their contributions. UNU reserves the right to edit, decline, or remove submissions in line with the purpose and values of the series. 

For inquiries, please contact: AINetwork@unu.edu

The image most policymakers still carry of an information threat is a single convincing fake — one doctored clip, traced, debunked, and removed. That picture is already out of date. The harder problem now is not the artefact but the system producing it: networks of automated accounts that no longer wait for a human to write each post, but plan, generate, test, and adapt their own messaging in pursuit of a goal. The unit of concern has shifted from the fake to the operation, and from the operation to the autonomous capacity to run thousands at once. 

This matters because such capability does something subtler than deceive. It manufactures the appearance of consensus. When coordinated agents flood a debate with what looks like organic agreement — adjusting tone by region, language, and platform in real time — the damage is not that any single claim is false. It is that the public can no longer read the room. A health ministry cannot tell whether opposition to a vaccination campaign is genuine community concern or a synthetic groundswell. A local official cannot tell whether the anger aimed at them reflects constituents or a rented crowd.  

Information integrity has become a question of whether a society can still perceive itself accurately. 

After working at the intersection of artificial intelligence and information manipulation for the better part of a decade, I have come to see one framing choice as among the most consequential in current AI governance: whether we file this under national security or under sustainable development. We have largely chosen the former, leaving the Sustainable Development Goals exposed. 

Security Isn't the Whole Story 

Treating automated influence operations as national security matters captures something true: these capabilities are used to interfere in elections, inflame conflict, and probe the resilience of states. But the security frame carries assumptions that quietly distort the response. It centers acute, attributable incidents over chronic erosion. It privileges detection and takedown over the slower work of sustaining trust. And it concentrates resources in well-resourced states with mature institutions — precisely the places that need it least. 

The harms that matter most for sustainable development are rarely the dramatic ones. They are cumulative and ambient: a persistent automated undercurrent that makes a childhood immunization drive look contested when it is not; coordinated amplification that erodes confidence in a local result until communities stop believing any outcome; adaptive narratives that reframe a climate adaptation program as a foreign imposition. No single post crosses a threshold worth a headline. The aggregate quietly degrades the institutional trust the SDGs depend on. 

Why This Is an SDG Issue 

Sustainable Development Goal 16 calls for effective, accountable, and inclusive institutions. That goal is unreachable in an information environment where the cost of manufacturing a plausible groundswell has collapsed toward zero while the cost of verifying what is genuine has barely moved. Earlier manipulation was bounded by human effort — someone had to write each message and run each account. Automation has removed that ceiling: a single operator can now direct systems that generate, localize, and continuously adapt campaigns across many languages and platforms at once. That asymmetry between fabrication and verification is the real story, and it is a development story before it is a security one. 

The consequences land hardest where institutions are already fragile, independent media is thin, and a single language community may have almost no automated safety tooling protecting it. Detection systems that perform well in English or Mandarin frequently fail in the languages spoken by the people most exposed — and automated operations are increasingly capable of working fluently in exactly those underserved languages. The result is a quiet inequity: the communities with the least institutional resilience receive the least protection, and are the easiest to flood. 

What Detection Cannot Fix 

There is a strong temptation to believe this is an engineering problem awaiting an engineering solution — that better classifiers will eventually catch the fakes. Practitioners know better. Detection is necessary, but it is structurally always a step behind generation, and that gap widens when the adversary is an adaptive system that can test what evades scrutiny and adjust faster than any review cycle. More to the point, detection does nothing to repair trust once it has been worn down. A society does not become resilient because its platforms remove more content. It becomes resilient because its institutions stay credible enough that a manufactured consensus finds no purchase. 

That points toward interventions the security frame tends to underfund: 

• Treating institutional credibility — transparent public communication, independent media, and verification capacity — as core development infrastructure, not an optional extra. 
• Building safety and detection tooling for low-resource languages, treating linguistic coverage as a matter of equity rather than commercial priority. 
• Designing for human judgement in the loop, since the goal is not a perfect filter but institutions that can absorb manipulation without fracturing. 
• Measuring trust, not just takedowns, so success is defined by institutional resilience rather than content removed. 

A Reframing, Not a New Budget Line 

The shift I am arguing for is conceptual before it is financial. When information integrity is filed under national security, it competes for attention with borders and cyberattacks, and it is governed by institutions built for secrecy rather than public trust.  

When it is understood as development infrastructure — as foundational to functioning institutions as clean water or reliable courts — it becomes a shared, cross-regional priority that development actors, multilateral bodies, and national governments can own together. 

This is not an argument for abandoning the security response. It is an argument for refusing to let it be the whole response. The societies that will weather the coming decade of automated influence are not necessarily those with the best detection algorithms. They are the ones whose institutions remain trustworthy enough that a manufactured consensus changes nothing important. Building toward that outcome is development work. We should fund it, govern it, and measure it as such. 

Disclosures 

Conflict of interest: The author declares no conflict of interest. 

Sensitive or controversial issues: This article addresses automated influence operations and their effects on electoral and institutional trust, which may be politically sensitive in some contexts. It does not reference any specific ongoing case or actor. 

Use of AI tools: AI tools were used for copyediting purposes and for the research of references.  

References 

United Nations. (2024). Global Digital Compact. New York: United Nations. 

United Nations. Sustainable Development Goal 16: Peace, Justice and Strong Institutions. https://sdgs.un.org/goals/goal16 

UNESCO. (2023). Guidelines for the Governance of Digital Platforms. Paris: UNESCO. 

OECD. (2024). Facts not Fakes: Tackling Disinformation, Strengthening Information Integrity. Paris: OECD Publishing.