The Economist’s latest framing of AI’s resource problem is not about compute or energy. It is about the courts. In a piece published August 6, the magazine applies Garrett Hardin’s 1968 tragedy of the commons to a startling new domain: Britain’s employment tribunals, where interim relief applications have surged a hundredfold as workers use ChatGPT or Grok to draft legal claims for free instead of paying lawyers.
The numbers are stark. Claims rose 39 percent in the year through March 2026, and the backlog jumped 55 percent to 64,000 unresolved cases, according to a memo from tribunal presidents Barry Clarke and Susan Walker. Many AI-generated filings run hundreds of pages, packed with fabricated laws and unrealistic demands. The Economist calls it “a tragedy of the commons, AI edition,” and the label fits: every individual worker gains from firing off a cheap, AI-drafted claim, but the collective result is a clogged system where genuine grievances wait longer for justice.
The courtroom is not the only commons under strain. The same logic applies to the data that trains the models drafting those claims. The AIToolly analysis of the Economist’s piece lays out the core mechanism: the public internet is a finite pasture, and every AI company has an individual incentive to scrape as much human-generated data as possible, while the collective effect is depletion and degradation. The “grass” of human data is being replaced by “synthetic weeds” of AI-generated content, a feedback loop that threatens model collapse.
The legal flood is the sharpest example yet of what happens when the tools of the commons are turned on the commons itself. Workers are using AI to draft claims, and the claims are so voluminous and so often fabricated that the system cannot cope. The memo from Clarke and Walker describes filings that are “unrealistic” in their demands and “fabricated” in their legal citations. The courts are now the overgrazed pasture, and the AI-generated filings are the synthetic weeds choking it.
The timing could not be worse. Labour’s new Employment Rights Act adds roughly 25 new grounds for claims and removes compensation caps, according to The Decoder’s coverage. That means the flood of AI-drafted claims will likely accelerate, not recede. Employers now pay more to respond to every claim, whether legitimate or made up, and workers with real grievances wait longer for a hearing. The tragedy is that the very tool meant to democratize access to justice is degrading the justice system for everyone.
The US faces a parallel crisis. One federal judge has called AI-generated lawsuits an “existential threat to the federal courts.” The pattern is identical: cheap drafting, voluminous filings, and a judiciary drowning in paper. But a study from Pakistan shows the outcome is not inevitable. Judges equipped with AI tools and training processed more cases, faster. The difference is not the presence of AI; it is whether the system adapts to filter and triage AI-generated filings rather than letting them pile up unchecked.
The deeper lesson for the AI industry is that the commons problem is not just about data. It is about the externalities of cheap generation. When the marginal cost of producing a legal claim, a news article, or a training dataset approaches zero, the incentive structure shifts. Every individual actor benefits from producing more, but the collective resource, whether a court docket or a training corpus, degrades. The Economist’s framing is a warning that the tragedy is not hypothetical; it is happening now, in real courts, with real backlogs.
The data commons is already showing the same strain. The AIToolly analysis notes that scaling laws tie model performance directly to the volume of high-quality training data. As models reach the limits of available text on the internet, competition for the remaining “pristine” data intensifies. Wikipedia, digitized books, and major news archives are being ingested by every major developer, and the marginal utility of that data is shifting. The overgrazing phase is here.
The pollution phase is accelerating too. As the internet becomes saturated with AI-generated text and images, future models will train on the output of their predecessors. Model collapse occurs when an AI model loses its grasp on reality or linguistic nuance because it has been trained on too much synthetic data. The errors, biases, and hallucinations of one generation are amplified in the next, erasing the ground truth that human data provides. The pasture is not just depleted; it is poisoned.
The industry’s response is already reshaping the landscape. Content providers, recognizing the value of their data, are implementing anti-scraping technologies, erecting paywalls, and seeking legal recourse. This is the “enclosure” phase, a historical parallel to the enclosure of common lands in England. The shift from an open internet to a fragmented landscape of proprietary data silos will increase the cost of AI development and favor large incumbents with the capital to purchase exclusive data rights.
The legal flood is a preview of what enclosure will look like in practice. When the commons is overgrazed, the response is to fence it off. The courts are already fencing themselves off, with tribunals imposing page limits and stricter rules on AI-drafted filings. The data economy will follow the same path, with premium information locked behind paywalls and restrictive licenses. The open internet that enabled the first wave of AI progress is closing.
There is a path forward, but it requires a new social contract between AI developers and the human creators who provide the foundational “nutrients” for the digital ecosystem. The AIToolly analysis suggests a pivot toward data-efficient architectures, small-data learning, and the curation of highly specialized, high-fidelity datasets. The focus must shift from “more data” to “better data” and “smarter algorithms.”
The courtroom crisis shows what happens without such a contract. The workers flooding Britain’s tribunals are not villains; they are rational actors responding to incentives. The AI companies scraping the internet are not villains either; they are doing what scaling laws demand. The tragedy is structural, not moral. The question is whether the industry can build the institutions to manage the commons before it collapses entirely.
The next few months will tell. The Employment Rights Act takes effect, the backlog grows, and the tribunals will be forced to adapt. The data enclosure accelerates, and the cost of training rises. The AI industry faces a choice: treat the commons as a finite resource to be stewarded, or watch it be fenced off, polluted, and litigated into irrelevance. The courts are already showing which way the wind blows.