The New Knowledge Pipeline
How AI, Philanthropy, and “Evergreen” Media Are Reshaping Intellectual Labor
The popularization of artificial intelligence is reshaping not only technology, but also the economics of knowledge production. Alongside entertainment and opinion, increasingly valued internet content is structured, explanatory, continuously updated information which can easily be parsed by AI systems. In this emerging landscape, projects modeled after Wikipedia occupy a particularly strategic position. They produce organized human knowledge in a form ideal for machine learning. While AI companies benefit from these digital ecosystems, who else benefits alongside them?
Wikipedia is perhaps the clearest example of this transformation. The Wikimedia Foundation has received donations from major tech firms including Google, Microsoft, Apple, GitHub, Adobe, and others. More recently, Wikimedia formalized commercial relationships with AI companies through Wikimedia Enterprise, licensing structured access to its data to firms including Microsoft, Meta, Amazon, Perplexity, and Mistral AI. "Wikimedia Enterprise isn’t just about getting tech companies to pay for their use; it also provides them access to Wikimedia projects at a volume and speed designed to meet their data needs."
One interpretation is that Wikipedia has become infrastructure for the AI economy. Large language models depend heavily on massive quantities of clean, human-written, frequently updated information, and Wikipedia is among the most trusted and machine-readable repositories ever created. Wikimedia officials themselves have acknowledged the pressure AI scraping places on their systems. "Wikipedia knowledge for AI training has driven up server demand and, subsequently, costs..." As such, Wikimedia is attempting to secure compensation for value that AI companies were already extracting for free. And its partnerships may represent an effort to force redistribution within an already unequal ecosystem.
Open collaborative knowledge production has become economically indispensable to AI systems.
The Observatory project, launched by the Independent Media Institute, presents itself as a “wiki” devoted to explanatory, evergreen political and intellectual content. Rather than publishing static essays, it aims to create continuously updated guides that synthesize the work of scholars, journalists, and public intellectuals into accessible explanatory texts.
The model resembles Wikipedia not only stylistically but structurally. The content is modular, educational, searchable, evergreen, and optimized for accessibility rather than academic permanence. In practice, this means translating dense intellectual work into machine-readable explanatory language.
The project’s relationship to AI is reserved for conversations with potential contributors and writers. Contributors were told that AI language models favor exactly this kind of content — wiki style, continuously updated, explanatory, digestible prose — and that this visibility within AI systems constituted part of the exposure contributors would receive in exchange for their intellectual labor.
In addition to AI systems benefitting from this publishing style, AI discoverability appears to be consciously integrated into the intellectual production's rationale and value proposition.
Traditional scholarship depends on stable citations through fixed editions and archival permanence. Evergreen and living documents complicate that structure because texts can change over time. While authors may receive bylines, the work itself becomes fluid. Citations become unstable, and the historical durability traditionally associated with intellectual production weakens. Not to mention that AI companions seldom cite their sources.
This creates an asymmetry. AI systems benefit precisely from dynamic, continuously updated explanatory content. But for scholars and writers, the professional value of unstable, constantly rewritten texts may be considerably lower than that of fixed publications with durable citation value, and whose readers are conscious of who they are reading. Or, simply, conscious. Which is why the labor issue becomes even sharper when payment is replaced with promises of exposure.
Exposure through AI systems is fundamentally different from traditional readership. Large language models often synthesize information without clearly preserving attribution or directing users back to original authors. Even when AI systems indirectly amplify ideas, the resulting economic value is captured primarily by technology platforms rather than by the intellectual workers whose labor improved the models’ outputs. This publishing structure aligns closely with the needs of the AI industry.
The funding ecosystem surrounding the Independent Media Institute further complicates the picture. Nonprofit databases show that IMI has received substantial funding through donor-advised philanthropy structures, namely the GS Donor Advised Philanthropy Fund, formerly known as the Goldman Sachs Philanthropy. Donor-advised funds have become increasingly influential mechanisms within elite philanthropy because they allow wealthy donors to receive immediate tax benefits while shielding the flow of donations from public scrutiny.
That same year, an IRS disclosure error unintentionally revealed several major contributors to the Goldman Sachs Philanthropy Fund, including Steve Ballmer (former Microsoft CEO), Laurene Powell Jobs (with stakes in Apple and Walt Disney), and Jan Koum (WhatsApp co-founder). The opacity is not accidental; concealment is one of the institutional features donor-advised funds explicitly provide. As a result, the inability to establish direct donor intent cannot automatically be treated as evidence that no structural alignment exists.
Brand new systemic incentives are coming to life. The AI economy is rewarding the production of simplified, structured, evergreen intellectual content optimized for machine absorption. Nonprofit media projects, open knowledge initiatives, philanthropic capital, and AI corporations all operate within this same ecosystem, even as their motivations may differ.
What emerges is a new political economy of intellectual production. Scholars, journalists, and public thinkers may increasingly be encouraged to reshape their work into AI-friendly formats not because such formats necessarily best serve human readers, but because they maximize discoverability, circulation, and machine usability — an offspring of the SEO friendliness of the early 2000s. Meanwhile, the greatest economic gains may accrue not to the producers of knowledge, but to the companies building commercial AI systems atop that knowledge infrastructure.
It is yet to be revealed whether intellectual labor in the AI era will remain a largely uncompensated public resource, or new models of ownership, compensation, and attribution will emerge before the infrastructure of human knowledge is permanently absorbed into machine economies.
Mirna Wabi-Sabi
Mirna is a Brazilian writer, editor at Sul Books and founder of Plataforma9. She is the author of the book Anarcho-transcreation and producer of several other titles under the P9 press.