Dispatch #12 — Critical Information Infrastructures in the Age of AI
Dispatch 12 — Critical Information Infrastructures in the Age of AI
There is a growing tendency to discuss artificial intelligence as if it were a software revolution. But the deeper story is infrastructural. AI is becoming a layer that reorganises labour, political communication, scientific production, media distribution and even moral authority. The central question is therefore who controls the infrastructures on which societies increasingly depend: compute, data, verification systems, satellite networks, trusted repositories, democratic communication channels and the information flows that connect them.
This dispatch follows that logic. Across regulation, elections, journalism, open-source intelligence and platform ecosystems, the same pattern emerges repeatedly: societies are rediscovering the strategic importance of trusted intermediaries, resilient public-interest infrastructures and governance models capable of resisting both manipulation and industrial-scale synthetic noise.
For news agencies in particular, this debate is essential. In an environment increasingly flooded by AI-generated content, manipulated narratives and agent-driven distribution systems, trusted verification networks are part of democratic resilience itself.
I. On power and the political economy of AI
Signal 12.1 – “When AI costs more than people”
Source: Fortune, “Nvidia executive says the cost of AI is greater than the cost of employees” –
Dispatch
A senior Nvidia executive argues that, at current scales, running advanced AI systems costs more than employing the people whose work they are meant to augment or replace. He points to the capital intensity of GPUs, energy and cloud infrastructure as a structural constraint on “AI everywhere” business models. The piece suggests that many firms underestimate these ongoing costs when they rush to automate or build AI features. It raises the question of who will ultimately be able to afford high-end AI, and who will be left buying cheaper, less reliable systems.
Reflections
• If the true cost of AI exceeds the cost of human labour in many cases, which kinds of work actually make economic sense to automate?
• What does it mean for newsrooms and news agencies deciding where AI belongs in their workflows — and where human judgement is both better and cheaper?
Signal 12.2 – “Banning AI-generated papers: arXiv’s first hard stop”
Source: 404 Media, “New arXiv rules on AI-generated papers ban”
Dispatch
This piece describes how arXiv, one of the main repositories for scientific research, is tightening rules around AI-generated content. The platform responds to a surge in low-quality machine-authored submissions that strain moderation and risk confusing readers about what has actually been peer-reviewed. New guidelines require clear disclosure of AI use and allow moderators to reject papers that appear to be largely synthetic or deceptive.
The debate mirrors concerns in journalism: when volume explodes and barriers to production collapse, quality assurance itself becomes infrastructure.
Reflections
• What practical standards should exist for disclosing AI assistance in research, journalism and policy documents?
• How can repositories and archives remain open and fast without becoming dumping grounds for AI-generated slop?
Signal 12.3 – “Brazil’s elections as an AI stress test”
Source: Tech Policy Press, “Brazil’s 2026 elections are its first real stress test for AI regulation”
Dispatch
The article argues that Brazil’s 2026 elections will be the first major test of the country’s new AI and platform regulations in a high-stakes democratic context. Lawmakers and regulators have put in place rules around political advertising, deepfakes and content moderation, but enforcement capacity and coordination with platforms remain open questions. Civil society groups warn that generative AI could turbo-charge disinformation campaigns, especially in encrypted and semi-closed ecosystems.
Brazil therefore becomes an early laboratory for one of the defining questions of this decade: can democratic governance mechanisms move faster than synthetic manipulation systems?
Reflections
• What should success look like for AI regulation in an election: fewer abuses, faster response, more transparency, or something else?
• What lessons from Brazil will be most relevant for European elections as AI tools become part of every campaign’s toolkit?
II. On governance, regulation and “living” AI policies
Signal 12.4 – “How much power does the EU AI Office really have?”
Source: Lawfare, “How much power does the EU AI Office actually have?”
Dispatch
Lawfare analyses the formal mandate and practical constraints of the new EU AI Office, which is tasked with overseeing implementation of the AI Act. On paper, the office coordinates enforcement, issues guidance and can investigate systemic risks, but many concrete powers sit with national regulators. The piece highlights uncertainties around staffing, expertise and political backing, which will determine whether the office becomes a genuine centre of gravity or merely a coordinating layer.
For industries like media and technology, this will shape how coherent and predictable AI oversight in Europe really becomes.
Reflections
• What capabilities, technical, legal and political , does an AI supervisor need to become an effective node in the regulatory network?
• How can news organisations engage with such bodies early, instead of only reacting once enforcement arrives at their doorstep?
Signal 12.5 – “AI guidelines as a living policy”
Source: WAN-IFRA, “The living policy: why AI guidelines are never finished” –
Dispatch
This WAN-IFRA piece argues that AI policies inside media companies cannot remain static documents but must evolve continuously as tools, risks and workflows change. The author warns against both rigid bans that quickly become obsolete and vague principles that provide no operational guidance.
Instead, AI governance should function as a living framework for experimentation, documentation, accountability and institutional learning.
Reflections
• How can organisations build feedback loops so that frontline experience with AI tools genuinely updates policy?
• In what ways can “living policies” be communicated externally to strengthen trust with audiences and partners?
Signal 12.6 – “When politicians abandon US messaging applications”
Source: Lawfare Media, “Politicians to ditch Signal for homegrown apps” –
Dispatch
Lawfare reports on a growing trend of politicians and public institutions moving away from global encrypted apps like Signal towards domestically developed messaging platforms. Officially, motives include security, sovereignty and integration with local infrastructures. Critics fear that such systems may also become easier to align with political interests or state access.
The broader issue goes far beyond messaging applications. Communication infrastructure itself is becoming geopolitical terrain.
Reflections
• Why are sovereign or bespoke messaging systems becoming strategically important in democracies?
• Which standards should apply for encryption, governance and oversight when such tools are used for public decision-making?
Signal 12.7 – “Journalism recognised as democratic infrastructure — but underfunded”
Source: European Federation of Journalists / AgoraEU
Dispatch
This article covers a European Parliament initiative explicitly framing journalism as democratic infrastructure deserving structural support rather than occasional subsidies. While symbolically important, journalist organisations argue that the proposed funding remains disconnected from the scale of the challenge.
The contradiction is revealing: societies increasingly recognise trusted journalism as essential infrastructure, while still financing it as if it were an optional market by-product.
Reflections
• What practical difference does it make when journalism is treated as infrastructure rather than merely a commercial product?
• How can funding mechanisms protect editorial independence while still creating long-term stability?
Signal 12.8 – “A papal encyclical on AI”
Source: OSV News, “Pope publishes encyclical on artificial intelligence”
Dispatch
OSV News reports that the Pope is publishing an encyclical dedicated to artificial intelligence, placing AI within a broader moral and social framework touching labour, inequality, dignity and concentration of power.
This matters because AI governance is no longer only a technical or regulatory discussion. It increasingly becomes a civilisational one.
Reflections
• How will a major religious text on AI influence public perceptions of acceptable technological use?
• Can shared ethical reference points across secular and faith-based institutions help anchor AI governance in more human-centred principles?
III. On journalism, trust and the business of news
Signal 12.9 – “Believability vs credibility”
Source: WAN-IFRA, “Believability vs credibility: what journalism can learn from creators”
Dispatch
This piece distinguishes between “believability” — the feeling that a voice is relatable and trustworthy — and “credibility”, based on evidence and institutional processes. Creator-journalism often excels at believability through proximity, tone and personality, while newsrooms emphasise credibility through verification and standards.
The author argues that journalism needs to better integrate both dimensions: making processes more visible and human-scaled without sacrificing rigour. Otherwise, audiences may increasingly gravitate toward believable voices that are less grounded in facts.
Reflections
• How can news organisations show their editorial processes in ways that increase both believability and credibility?
• When does institutional distance remain an asset, even if it feels less immediately “believable”?
Signal 12.10 – “Generation AI and the news”
Source: Common Sense Media, “Generation AI” report
Dispatch
The “Generation AI” study explores how children and teenagers interact with AI tools, platforms and content, including how they encounter news and information. It finds that many young people use AI systems for homework help, explanations and even current affairs, often without clear understanding of how these tools work or what their limitations are.
The report highlights concerns around bias, privacy and exposure to misinformation, but also documents a pragmatic openness among youth to using AI as a learning aid. For newsrooms, it underlines the importance of designing products and educational approaches that meet this generation where it already is.
Reflections
• What does news literacy look like in a world where asking an AI system becomes a default way to understand events?
• How can news organisations build youth-specific interfaces and experiences that integrate AI in more transparent and educational ways?
Signal 12.11 – “AI slop, backlash and detection”
Source: What’s New in Publishing (Substack), “The AI slop backlash is good news” –
Source: NewsGuard, “Real-time AI content-farm detection datastream”
Dispatch
One article argues that the wave of low-quality AI-generated “slop” flooding the web may ultimately benefit quality publishers by making audiences more alert to the difference between junk and journalism. In parallel, NewsGuard announces a real-time datastream designed to detect AI content farms, giving advertisers, platforms and publishers signals to avoid polluted inventory.
Together, these developments suggest the early formation of an “anti-slop ecosystem”: user backlash, detection tools and premium brands converging to revalue verified reporting.
Reflections
• What role should independent rating and detection services play in the economics of advertising and distribution?
• How could a shared clean datafeed for news, created by news agencies, become a competitive advantage in AI-driven environments?
Signal 12.12 – “AI agents and liquid content”
Source: CJR Tow Center, “AI agents are coming for news – can publishers reclaim control?”
Source: Journalism.co.uk, “Preparing for the second act of AI: agents, liquid content and AI deals”
Dispatch
These pieces map a near future in which AI agents — not human readers — become the primary consumers of news APIs, repackaging and routing content for end users. They describe “liquid content” flowing through multiple agentic systems, from personal assistants to enterprise dashboards, often far removed from the original publisher environment.
The articles ask whether publishers can shape this landscape through standards, licensing and technical controls, or whether they risk becoming invisible suppliers inside an agent-dominated ecosystem.
For news agencies in particular, this future is simultaneously a threat and an opportunity: the possibility of becoming trusted source layers for AI systems themselves.
Reflections
• What technical and contractual levers do news agencies need in order to remain sustainable and independent?
• How will editorial standards and attribution function when most “readers” are software systems acting on behalf of humans?
IV. On critical information infrastructure and alternative ecosystems
Signal 12.13 – “Mapping the blind spots from space”
Source: Columbia Journalism Review
Dispatch
CJR examines how satellite imagery and open-source intelligence have become indispensable tools for covering conflicts and closed regimes, while also revealing their own blind spots. Companies like Planet and Maxar provide near-real-time imagery that journalists and researchers use to document troop movements, damage and infrastructure.
But the piece also demonstrates that visibility remains uneven: some crises are extensively documented, while others remain effectively invisible because of technical, commercial or political constraints.
Satellite systems are therefore becoming part of global information infrastructure — powerful, but never fully neutral.
Reflections
• What are the risks when access to strategic imagery depends on a handful of private actors?
• How could public or cooperative satellite initiatives improve transparency around crises that currently remain in the shadows?
Signal 12.14 – “Digital gardens beyond Big Tech”
Source: LatAm Journalism Review
Dispatch
This piece recounts how community media outlets in Latin America are building “digital gardens”: smaller, locally controlled ecosystems designed to reduce dependence on major platforms. They invest in newsletters, websites, messaging groups and offline communities in order to cultivate slower but more resilient audience relationships.
The strategy reflects more than platform frustration. It reflects a desire to realign infrastructure with mission instead of outsourcing distribution to systems governed by entirely different incentives.
Reflections
• What can larger news organisations learn from community media about building smaller, more intentional digital spaces?
• How could news agencies support such ecosystems through feeds, tools or training?
Signal 12.15 – “Virtual assistants, AI and who really writes LinkedIn”
Source: Rest of World
Dispatch
Rest of World shows how a growing industry of low-paid virtual assistants in the Philippines use AI tools to write LinkedIn posts and comments for Western executives, effectively running a “thought leadership” content mill for around seven dollars an hour.
These assistants are hired to understand engagement logic, maintain constant online presence and manufacture professional authenticity — often with little disclosure that neither the text nor the engagement originates from the named individual.
Meanwhile, LinkedIn simultaneously tightens restrictions against third-party automation and scraping while continuing to permit selected API-based ecosystem partners.
The result is a striking paradox: authenticity itself becomes industrialised, outsourced and partially synthetic.
Reflections
• What does “authenticity” mean on professional networks when much of the engagement is produced by offshore assistants and AI systems rather than the named individuals themselves?
• Which transparency standards should apply once AI-assisted ghostwriting becomes a normal cross-border service industry?
Signal 12.16 – “The Hague signal: news agencies as democratic infrastructure”
Source: EANA / European Alliance of News Agencies
Dispatch
At the EANA Spring Conference 2026 in The Hague, hosted by ANP under the title The World of Disinformation: Challenges & Solutions for News Agencies, European news agencies treated disinformation not as a side issue but as a structural challenge for democracies and public governance.
Speakers such as Svitlana Slipchenko (VoxUkraine) demonstrated how information manipulation has become a permanent geopolitical instrument aimed at weakening democratic resilience, while Saman Nazari (Alliance4Democracy) presented a genuine whole-of-society response linking OSINT, media monitoring and institutional cooperation.
AFP fact-checking practices, GPT-NL initiatives, multimedia verification workflows and the EANA disinformation questionnaire all pointed in the same direction: news agencies are quietly building operational capabilities to detect, verify and contextualise false narratives at scale — not only for media clients, but for the broader democratic ecosystem itself.
The Hague therefore crystallised a deeper conclusion running through this dispatch: trusted news agencies are evolving from media organisations into critical democratic information infrastructure.
Reflections
• If news agencies are recognised as democratic infrastructure, what long-term obligations does this create for states, institutions and commercial actors in terms of support, access and legal protection?
• How can agencies operationalise a whole-of-society response to disinformation while preserving editorial neutrality and independence?
Conclusion — From media companies to infrastructure
One theme cuts through almost every signal in this dispatch: trust is becoming infrastructural.
Satellite constellations. Verification networks. OSINT communities. Messaging systems. AI governance bodies. Detection layers against synthetic content. Trusted archives. Human-centred editorial workflows. These are no longer peripheral tools around democracy. They increasingly are democracy’s operating system.
In that environment, the role of news agencies changes fundamentally.
Historically, agencies were often perceived as wholesalers of information operating quietly behind the visible media landscape. But in an AI-driven ecosystem saturated by synthetic content, fragmented narratives and agentic distribution systems, the ability to provide verified, neutral and resilient information flows becomes strategically essential.
The parallel with the EANA visit the European Space Agency is not accidental. Space infrastructure quietly enables navigation, communications, climate monitoring and security without most citizens ever directly seeing the systems behind it. Trusted information infrastructures increasingly play a similar role inside democratic societies: invisible when functioning well, existential when weakened.
The challenge for Europe is therefore not only technological competitiveness. It is institutional resilience.
The question is no longer whether trusted information infrastructures matter. The real question is whether democratic societies will recognise their strategic value early enough to preserve them before synthetic abundance makes trust itself a scarce resource.
Photo: Participants of the EANA Spring Conference 2026 at ESA ESTEC in Noordwijk, hosted by ANP. The setting itself carried symbolic weight: Europe discussing disinformation, verification and democratic resilience inside one of its most important scientific and infrastructural institutions.
Copyright ANP.