Stefaan Verhulst

Report by Anna Colom, Elena Murray, and Marta Poblet: “On the diagnosis side, our review identifies multiple, interconnected challenges. Namely: Generative AI providers scrape journalistic content at massive scale while returning negligible traffic or compensation. AI bot traffic also imposes disproportionate infrastructure strain, which is even harder for smaller organisations to cope with. Meanwhile, although exact figures on declining traffic vary, ‘zero-click’ searches are diverting audiences away from publishers. Together, these trends endanger the business models of news organisations, already walking on thin ice following the platformisation initiated by social media companies. Likewise, opacity in Generative AI models and outputs, inherent errors in both content accuracy and attribution, the limitations of Generative AI to summarise journalistic content in context-specific and nuanced ways, and the bypassing of original sources and editorial gatekeepers risk undermining the integrity of information as a key pillar for democracy. The growing concentration of Generative AI power in a few hands, as well as concentration in ways information is generated (with English-centric and large national media favoured over more local or diverse sources), further erodes pluralism, equity, and diversity…(More)”.

Field Guide prepared by Denice W. Ross and Christopher Steven Marcum: “The purpose of this guide is to provide a more complete context for federal data users and stakeholders that will inspire them to consider a broader range of data types in their research and advocacy; we also hope it will also inform national dialogues about the future of federal data.

Whatʼs included? The Guide is organized into eight primary categories of federal data (described on the
right), each representing distinct collection methods, policy frameworks, and use cases. This field guide focuses primarily on publicly available datasets created, maintained, and published by executive branch agencies of the federal government. This Guide does not include sensitive or classified datasets, or
derivative works such as reports or interactive web tools that use data…(More)”

Guidance Note by the Council of Europe: “Providing for unimagined opportunities, at scale and at speed, Generative AI also presents growing risks for freedom of expression and democratic processes, including with regards to the fragmentation of the information space through hyper-personalised experiences, and the lack of transparency, accuracy, repeatability, reliability, and the potential for bias and manipulation, of AI-generated content. The Guidance Note addresses these issues looking into the specific impact on the right to freedom of expression, enshrined in Article 10 of the European Convention on Human Rights.  

Firstly, it outlines the key characteristics of Generative AI technology and its lifecycle, by providing a shared vocabulary and offering a compass for its analysis. Then, the Guidance Note explores how Article 10 of the European Convention on Human Rights and the case-law of the European Court of Human Rights can guide the protection of freedom of expression in the context, and across the lifecycle, of Generative AI. Thirdly, it identifies the structural implications that, both at an individual and societal level, affect the foundations of freedom of expression. Standardisation of expression, hallucination, deep fakes, voice cloning, disinformation and opinion manipulation, being only some of known use cases. Finally, to ensure that Generative AI applications, their design and use, uphold and promote freedom of expression, the document delivers a concrete set of actionable measures for policymakers and other relevant stakeholders, through an agile governance cycle built on four interlocking areas: observe, assess, enable and empower…(More)”.

Framework by Helen McElhinney, Anjali Mazumder, Michael Tjalve, Suzy Madigan, and
Sarah Spencer: “SAFE AI is governance infrastructure for humanitarian AI. It guides organisations through a four-stage Implementation Journey from problem definition to deployment and monitoring, applies a three-tier risk classification with proportionate obligations at each tier, and sets formal Decision Gates at each stage where progression is tested against humanitarian principles, protection requirements, and responsible-refusal conditions.

It does this through a set of named tools deployed at specific points in the lifecycle:

• SAFE AI Onboarding and Readiness Checklist: at problem definition, to establish whether the conditions for responsible AI use exist.
• SAFE AI Impact Assessment: at the first and second Decision Gates, to test whether the use case should proceed.
• SAFE AI Architecture and Procurement Guides: at design and procurement, to secure right-to-audit, model change notification, data ownership, and exit conditions before deployment.
• SAFE AI Technical Assurance: at development and ongoing, to verify performance against documented baselines.
• SAFE AI Transparency Card: the central governance record, documenting decisions, risks, and safeguards across the lifecycle…(More)”.

News release by Smart Data Research UK: “… launched its new data catalogue at the Digital Footprints Conference, making it easier than ever for researchers to find and access smart datasets from across our six data services.

The catalogue (beta) brings together datasets covering finance, energy, transport, health, imagery and more in one searchable place. Researchers can filter by data type, theme, and access conditions, then follow direct links to access the data.

Featured datasets include Zoopla property rental listings, synthetic electric vehicle charging session data, and Spend Dynamics — microdata on household expenditure and consumer finances. Search the catalogue…(More)”.

Article by Krzysztof Pelc: “We are good at predicting what machines will do better; we are far worse at predicting what people will value differently once that happens. Yet the history of technological disruption suggests a fairly consistent pattern: when one property becomes abundant, perceived value migrates elsewhere. The Arts and Crafts aesthetic thus rose up as a challenge to factory production. Similarly, despite quartz watches making accuracy trivial by the 1970 s, mechanical watches once more dominate the global market by value (Raffaelli 2019). Technological shocks alter not only the goods on offer, but also the basis by which those goods are evaluated. What had seemed central is downgraded; what had been incidental becomes precious.

The direction of this change is invariably towards the human—not out of sentiment, but because in the wake of technological shocks, it is the human aspect that grows distinctive. The advent of large language models (LLMs) is beginning to have a similar effect on all writing. LLMs make verbal fluency cheap, they make competent prose abundant. As that happens, the old premium on flowing prose will weaken. If smoothness can be summoned on demand, smoothness no longer distinguishes. The scarce good will no longer be fluency, but provenance: whether a text can be traced to a particular human sensibility, lived experience, and intention. Call it the flight-to-humanity effect.

Usually, this revaluation works to the humans’ advantage. It’s the phenomenon that protects the radiologist, the craftsperson, the live performer, once their core output is superseded by machines. But writing will likely be an exception.

The problem is that writing is peculiarly ill-suited to certifying its own origins. In most domains, human provenance remains legible in the thing itself. A handmade bowl can bear the mark of its maker; a performer is present in the act; a physician’s judgment is tied to the physical person and their credentials. Writing is different. It arrives as a finished product, stripped of the conditions of its making. The reader sees the result, but not the process that produced it. Novels, essays, love notes, wedding speeches: none carry intrinsic evidence of authorship. This is not simply a practical difficulty. It’s a property of writing itself. And it’s what means that suspicion, once introduced, extends to every text alike…(More)”.

Review by Blanton Alspaugh: ““Where are we going, and why are we in this handbasket?”

Bumper-sticker theology poses the question. Paul Kingsnorth offers an answer in his prophetically pitched new book, Against the Machine: On the Unmaking of Humanity. Kingsnorth, a British philosopher and author living in the west of Ireland, joins a line of trenchant predecessors ranging from Paul the Apostle and G.K. Chesterton to René Guénon and Jacques Ellul. With an unflinching look at what we have made of the world, he sees “a machine made of human parts,” as Lewis Mumford characterized it. The Machine is the implacable, disembodied culmination of humanity’s fatal craving for knowledge and power. It respects no boundaries and obeys no law but its own. It advances more rapidly than our ability to manage it. Its origin is spiritual, and its consequences are eschatological. And Kingsnorth believes we must find a way to resist it.

As Kingsnorth has it, the West and Christendom are synonymous, and we are living through their death. But these terms are contested. “The West” for liberals is the Enlightenment and all that proceeded from it—“parliamentary democracy, human rights, individualism, freedom of speech.” For conservatives, it is a blend of cultural values—the traditions of “family life, religion and national identity, and…capitalist economics.” And for postmodern leftists, the West is little more than “a front for colonization, empire, racism.” As for Christendom, we may stipulate that it has often been very un-Christian, but it has still, as Christopher Dawson explained, been “the Christian Church, which provided an effective principle of social unity.” Unmoored from the sacred order, we live now among what Kingsnorth calls the “beautiful ruins” of Christendom.

Kingsnorth places the origin of the Machine in the biblical Fall, when we first saw something to be desired but were forbidden to take it. We took it anyway, and Kingsnorth makes a strong case that the Machine is in fact our ongoing project to become like gods. It is the accumulation of our efforts to transcend all boundaries, push past all limits, and overturn all traditions. The logic of the Machine is growth for its own sake; control and efficiency are its primal impulses; and the only value it recognizes is money, the market, Mammon. Money is how we obtain what we want, and the ceaseless stream of things to want is what the Machine uses to ensnare us. But, cosmically speaking, this money is play money—we are born with none, and we take none with us when we die. The true cost of obtaining what we want—going right back to that forbidden fruit in the Garden—is alienation, disenchantment, and dehumanization…(More)”.

Article by Ananya Bhattacharya: “Last month, South Africa withdrew its Draft National Artificial Intelligence Policy 17 days after it was published because the document cited fake research, created by AI.

The incident tarnished a historic moment, as South Africa was set to become the first African nation to adopt a policy establishing a formal ethics board to oversee AI outside the West. “The most plausible explanation is that AI-generated citations were included without proper verification,” Solly Malatsi, South Africa’s minister of communications and digital technologies, wrote in a statement. “There will be consequence management for those responsible for drafting and quality assurance.”

This is the first time a government has withdrawn a document over AI hallucinations, but certainly not the first time AI hallucinations have appeared in official materials. AI-generated text or citations have slipped into official or quasi-official documents several times, raising concerns about accountability and highlighting the need for human verification…(More)”.

Paper by Scott A. Brave, Erin E. Crust, Stefano Eusepi, Bart Hobijn & Ayşegül Şahin: “Interpreting real-time labor market conditions is challenging because commonly used indicators are noisy, revised over time, and often send conflicting signals. In practice, policymakers and market participants describe labor market developments using a shared narrative language centered on labor demand, labor supply, and matching frictions. In this paper, we show that empirical measures of these narrative concepts can be recovered from latent factors that summarize the joint movements of a broad set of high-frequency U.S. labor-market indicators. We use ninety-four labor-market indicators, over the period from 1960 to 2026, and construct measures for labor demand, long-run labor supply, short-run labor supply, and matching efficiency by selecting the factors that satisfy a limited set of restrictions on how underlying forces map into observed data. We find that labor demand and short-run labor supply account for most of the common variation in labor-market indicators. Our results also show that assigning narrow interpretations to individual indicators can lead to misleading conclusions about underlying labor market conditions. Applying the framework to the post-pandemic period reveals that although labor demand recovered briskly after the acute phase of the pandemic, it cannot account for the large rise in vacancies and quits. Instead, movements in short-run labor supply and matching efficiency play a central role. We also show that the “soft-landing” episode from 2023 through 2025 was characterized by a joint decline in labor demand and short-run labor supply, which slowed payroll growth while generating only a moderate increase in the unemployment rate…(More)“.

Article by Sara Radin: “For years, the internet sold us the idea that connection doesn’t have to be local to be meaningful. Your people could live anywhere: in a Discord server, a group chat of far-flung friends, or a TikTok comment section. Geography was optional.

Now, more people are turning toward the ones physically closest to them: the neighbor down the block, the parent from the playground, the person whose wifi shows up in your network list. It’s not just about wanting connection; folks are looking for support. Childcare is expensive. Rent and groceries are high. Climate emergencies are more frequent. For many Americans, the difference between stability and crisis comes down to whether someone nearby can help.

Call it neighborism: the growing practice of treating proximity as a resource. Increasingly, digital tools aren’t replacing local relationships — they’re helping activate them.

Sometimes it looks small: introducing yourself to the people on your floor, starting a group chat for your building or block, sharing babysitters, watering a neighbor’s plants. But it can also look overtly political.

In Minneapolis, community responses to ICE activity blurred the line between everyday care and organized resistance. As federal immigration enforcement ramped up this winter, residents organized patrols, filmed arrests, shared alerts, and trained one another to document potential abuses. What emerged was something bigger than “borrow a cup of sugar” friendliness. It was infrastructure: informal, fast-moving, and built on trust. And what happened there isn’t an outlier; it’s a large-scale example of a broader shift already underway.

Getting to know your neighbors isn’t new, but its visibility is. After decades of isolation and a slow drift toward digital, long-distance connection, people are embracing an old-fashioned idea: Communities function best when people feel responsible for one another…(More)”.