Introduction: The Commons Under Siege
The open knowledge movement was built on a radical proposition: that human knowledge, freely shared and collaboratively maintained, could become a global public good accessible to everyone. For two decades, that proposition held. Wikipedia became the world's largest encyclopaedia. Creative Commons licensed billions of works. Open data initiatives made government information accessible to citizens and researchers. The digital commons flourished.
Then came generative AI — and with it, a structural challenge that the open knowledge movement had not anticipated and is only now beginning to fully comprehend. The same openness that made the commons a global resource has made it the primary fuel for a new generation of proprietary AI systems that extract value at unprecedented scale, return little to the communities that created the content, and are systematically displacing the human interfaces through which the commons was accessed and sustained.
This timeline traces the evolution of that challenge — from the early warnings to the strategic responses now taking shape — and examines what the open knowledge movement's response tells us about the future of shared digital infrastructure in the age of artificial intelligence.
Phase One: The Extraction Era (2020–2023)
The First Knowledge Loops
To understand the current crisis, it helps to understand the history of how the digital commons has interacted with successive waves of internet technology. Wikimedia's analysts describe this history in terms of "knowledge loops" — cycles of content and resource exchange between the commons and the broader information ecosystem.
The first knowledge loop was defined by search engines. Google and its competitors indexed Wikipedia and other open knowledge resources, driving enormous volumes of human traffic to the commons. That traffic translated into donations, volunteer engagement, and the social visibility that sustained the movement. The relationship was reciprocal: search engines extracted value from the commons, but they also delivered value back in the form of human attention and financial support.
The second knowledge loop was defined by social media. Platforms like Facebook and Twitter amplified open knowledge content, again driving traffic and engagement. The relationship was more ambiguous — social media also spread misinformation that the commons had to work to counter — but the basic reciprocity held.
The Third Loop Breaks the Pattern
The third knowledge loop, driven by large generative AI models, broke this pattern of reciprocity. Beginning around 2020 and accelerating dramatically through 2022 and 2023, AI developers began ingesting the digital commons at scale. Wikipedia's structured data, Creative Commons-licensed images, open-access academic papers, and publicly available government datasets became foundational training material for models that would go on to generate billions of dollars in commercial value.
The critical difference from previous knowledge loops was the absence of reciprocity. Search engines sent humans to Wikipedia; AI systems answered questions directly, without attribution, without traffic, and without financial contribution. The commons was being used as infrastructure for proprietary systems that had no structural incentive to sustain it.
Phase Two: The Reckoning (2024–2025)
By 2026, AI systems had driven an 8% decline in human traffic to Wikimedia projects while bot activity surged 50% — the digital commons was becoming infrastructure for proprietary AI while becoming invisible to the humans it was built to serve.
The Traffic Collapse
By 2024, the structural impact of the third knowledge loop was becoming measurable. Wikimedia's internal data showed an 8 per cent decline in human traffic to Wikimedia projects, accompanied by a 50 per cent surge in bot activity. The pattern was clear: AI systems were consuming Wikimedia content at increasing rates while the human users who had historically sustained the movement — through donations, volunteer editing, and social engagement — were being displaced by AI interfaces that answered their questions without directing them to the source.
By 2026, AI systems had driven an 8% decline in human traffic to Wikimedia projects while bot activity surged 50% — the digital commons was becoming infrastructure for proprietary AI while becoming invisible to the humans it was built to serve.
The Wikimedia Foundation began describing this dynamic as the risk of becoming a "dark commons" — a foundational, high-quality data source for proprietary AI systems that had become structurally invisible to human users and was losing its ability to shape how knowledge was governed and consumed. The irony was acute: the more valuable Wikipedia's content became as AI training data, the less visible it became to the humans it was built to serve.
The Paradox of Open
The open knowledge movement began grappling with what Open Future and Wikimedia CH described as the "Paradox of Open": the commitment to free, non-discriminatory access, in an environment of extreme power asymmetry between large AI developers and volunteer-maintained commons, was facilitating extraction rather than promoting equity.
The 'Paradox of Open' defines the current crisis: the commitment to free sharing, in an environment of extreme power asymmetry, facilitates extraction by large-scale AI developers rather than promoting the equity the open movement was founded to advance.
This was not a failure of the open knowledge movement's values — it was a failure of the institutional and technical infrastructure that had been built to operationalise those values. The licences, norms, and governance structures of the open knowledge movement had been designed for a world of human users and human-scale reuse. They were not designed for a world in which a single AI company could ingest the entire Wikipedia corpus in hours and use it to build a commercial product worth billions of dollars.
Creative Commons Confronts the AI Era
Creative Commons, whose licences govern billions of openly licensed works, faced a parallel challenge. The CC licence suite — designed to enable sharing and reuse while preserving attribution and other creator rights — had not been designed with AI training in mind. A work licensed under CC BY (attribution required) could be used to train an AI model, but the model's outputs would not carry attribution to the original creator. A work licensed under CC BY-NC (non-commercial use only) might or might not be usable for commercial AI training, depending on how "non-commercial" was interpreted in the context of model training.
In 2024 and 2025, Creative Commons began a fundamental rethinking of its role in the AI era. The organisation recognised that its existing licence infrastructure was insufficient to address the new dynamics of AI-scale data reuse, and that a new layer of technical and governance infrastructure was needed.
Phase Three: Strategic Response (2026)
Wikimedia's 2030 Mission
The 'Paradox of Open' defines the current crisis: the commitment to free sharing, in an environment of extreme power asymmetry, facilitates extraction by large-scale AI developers rather than promoting the equity the open movement was founded to advance.
In early 2026, a white paper developed by Open Future and Wikimedia CH proposed a "Wikimedia and AI mission" for 2030, resting on three strategic pillars. The paper represented a fundamental shift from a defensive posture — trying to protect the commons from AI extraction — to an active, mission-oriented strategy that sought to shape how AI interacts with open knowledge.
The first pillar was pro-commons governance: establishing new norms for attribution, reciprocity, and ecosystem protocols to ensure that machine use of the commons is accountable and sustainable. This included advocacy for regulatory frameworks that require AI systems to attribute their sources, contribute financially to the infrastructure they depend on, and respect the governance norms of the communities that maintain open knowledge resources.
The second pillar was the Wiki AI technical stack: investing in the development of commons-aligned data and application layers, and contributing to the creation of open, public AI models. Rather than simply providing training data for proprietary AI systems, the Wikimedia movement would develop its own AI capabilities — tools that could help editors, improve content quality, and demonstrate that AI and open knowledge could be mutually reinforcing rather than structurally opposed.
The third pillar was mission network and investment: coordinating efforts across the movement and with external partners to ensure the necessary resources were available to maintain the commons. This included the expansion of Wikimedia Enterprise — the Foundation's commercial service for high-volume, real-time access to Wikimedia data — as a mechanism for ensuring that large-scale commercial reusers contribute financially to the infrastructure they depend on.
Wikimedia Enterprise: Commercialising Access to Sustain the Commons
Wikimedia Enterprise, launched five years ago, has emerged as a key tool for managing the pressures of the third knowledge loop. By providing a commercial service for high-volume, real-time access to Wikimedia data — with structured APIs, guaranteed uptime, and dedicated support — the Foundation creates an incentive for large-scale commercial reusers to pay for the data they consume rather than simply scraping it for free.
As of 2026, Wikimedia Enterprise has become a core component of the Foundation's sustainability strategy, with income capped at 30 per cent of total annual revenue to preserve the organisation's independence and non-commercial character. The model is carefully calibrated: free access for human users and small-scale reuse remains the default, while large-scale commercial reuse — the kind that AI companies engage in — is channelled through a paid service that funds the infrastructure that makes the commons possible.
The Mosaic Mission and European Data Commons
July 2026 saw the launch of the "Mosaic" mission — a consortium-led initiative to protect the Wikimedia ecosystem and broader knowledge commons from the structural impacts of generative AI. The mission brings together libraries, archives, research institutions, and open knowledge organisations to develop shared governance frameworks, technical standards, and advocacy positions for the AI era.
Parallel to the Mosaic mission, feasibility studies and policy briefs from 2025 and 2026 have proposed a "European Books Data Commons" — a multilingual, AI-ready dataset of book content governed by libraries for public-interest use. The proposal reflects a broader European strategy of building "Public AI" infrastructure: AI systems designed as public goods rather than commercial products, trained on openly governed datasets rather than proprietary scrapes.
Creative Commons: From Signals to Infrastructure
Creative Commons' strategic response to the AI era has crystallised around what the organisation calls the "CC Signals" framework — a new layer of technical and governance infrastructure designed to enable creators and data stewards to set terms for AI access to their content.
The framework moves away from the binary "open vs. closed" choices that characterised the original CC licence suite, towards a more nuanced "spectrum of participation" that allows creators to specify how their content can be used by AI systems. This includes carefully scoped AI opt-outs that preserve agency while protecting public-interest uses such as research, education, and preservation.
Creative Commons has declared that in the AI era, attribution is no longer just a preference but a foundational requirement — a shift from a culture of sharing to a culture of accountable sharing.
Creative Commons has declared that in the AI era, attribution is no longer just a preference but a foundational requirement — a shift from a culture of sharing to a culture of accountable sharing.
Central to the CC Signals framework is a new emphasis on attribution as a foundational requirement rather than a mere preference. Creative Commons has declared that AI systems — particularly Retrieval Augmented Generation (RAG) models, which can trace the provenance of their outputs — must provide attribution to the sources they draw on. This represents a significant shift in the organisation's positioning: from a culture of sharing to a culture of accountable sharing, where the benefits of openness are conditioned on the maintenance of the attribution norms that sustain the commons.
The Governance Questions That Remain
Who Pays for the Commons?
The most fundamental unresolved question in the AI-commons relationship is financial: who pays for the maintenance of the digital infrastructure that AI systems depend on? The open knowledge movement has historically been sustained by a combination of volunteer labour, small donations from individual users, and grants from foundations and governments. None of these funding mechanisms were designed to scale with the demands of AI-era data consumption.
Proposals for addressing this gap include AI levies on commercial companies — taxes on AI revenue or compute usage that would fund public knowledge infrastructure — and mandatory financial contributions from AI companies that use open knowledge resources above certain thresholds. The Open Future organisation has advocated for designating foundation models as "Core Platform Services" under digital market regulations, which would require mandatory third-party access to training data held by gatekeepers and could create mechanisms for financial reciprocity.
The Community Governance Challenge
Beyond the financial question lies a deeper governance challenge: how should the communities that maintain open knowledge resources make decisions about AI use of their content? The English Wikipedia community's 2026 decision to prohibit the use of LLMs to generate or rewrite article content — with limited exceptions — illustrates the complexity of these decisions. The policy reflects a genuine tension between embracing AI for productivity and maintaining the human-centric, collaborative nature of collective intelligence that has defined Wikipedia since its inception.
These governance decisions cannot be made by the Wikimedia Foundation alone — they require the engagement of the volunteer communities that actually create and maintain the content. Building governance frameworks that are both technically sophisticated enough to address AI-era challenges and democratically legitimate enough to command community support is one of the central challenges facing the open knowledge movement in 2026.
Conclusion: The Commons at a Crossroads
The open knowledge movement stands at a genuine crossroads. The digital commons that it has built over two decades — Wikipedia, Creative Commons-licensed works, open government data, open-access research — has become foundational infrastructure for the AI systems that are reshaping the global economy. That foundational role is both an opportunity and an existential risk.
The opportunity lies in the leverage that comes from being essential infrastructure: the open knowledge movement has a stronger negotiating position with AI companies, regulators, and funders than it has ever had before. The risk lies in the possibility that the movement fails to convert that leverage into sustainable governance frameworks and financial models before the structural damage from the third knowledge loop becomes irreversible.
The strategic responses taking shape in 2026 — the Wikimedia 2030 mission, the Mosaic initiative, the CC Signals framework, the European Books Data Commons — represent serious, sophisticated attempts to navigate this crossroads. Whether they will be sufficient depends on factors that extend beyond the open knowledge movement itself: the regulatory choices of governments, the commercial decisions of AI companies, and the willingness of the broader public to recognise that the digital commons is not a free resource but a maintained infrastructure that requires sustained investment to survive.
The third knowledge loop has broken the reciprocity that sustained the digital commons for two decades. Building a new reciprocity — one that is appropriate for the scale and dynamics of the AI era — is the defining challenge of the open knowledge movement in the years ahead.
