The Double Standard of Data Scraping: Aaron Swartz Was Prosecuted While Meta Walks Free

A Question of Justice Spanning Over a Decade

In 2011, a young programmer was indicted by federal prosecutors on 13 felony counts for bulk-downloading academic papers from the JSTOR database, facing up to 35 years in prison and $1 million in fines. His name was Aaron Swartz—co-author of the RSS specification, co-founder of Reddit, early contributor to Creative Commons, and a champion of the open knowledge movement. In January 2013, at just 26 years old, he took his own life while awaiting trial.

Aaron Swartz was one of the most influential early builders in internet history. At 14, he participated in drafting the RSS 1.0 specification—RSS (Really Simple Syndication) is an XML-based protocol that allows users to subscribe to website content updates, and it remains foundational infrastructure for podcast distribution and news aggregation to this day. He went on to co-found Reddit (now one of the world's largest community forums) and was deeply involved in designing the technical architecture of Creative Commons, a licensing framework that provides the legal foundation for open content sharing on the internet. He was also an early contributor to the Markdown format specification and a key organizer in the movement against SOPA (the Stop Online Piracy Act). His life trajectory was the purest embodiment of the internet's open spirit.

More than a decade later, when tech giant Meta was exposed for mass-scraping (and even obtaining through piracy channels) vast amounts of copyrighted books and data to train AI models, we see an entirely different picture: lawsuits abound but cause barely a ripple, business continues as usual, and the stock price keeps climbing. This stark contrast is precisely the core issue that recently sparked 886 upvotes and 204 comments on Hacker News.

Same Data Scraping, Radically Different Legal Fates

Aaron Swartz's "Crime": Downloading Academic Papers

What Swartz did was essentially bulk-download academic papers from JSTOR through MIT's campus network. Most of these papers were products of publicly funded research, and his intention was to return publicly funded knowledge trapped behind paywalls back to the public. Even though JSTOR itself subsequently dropped its civil claims, federal prosecutors persisted with criminal charges, invoking the Computer Fraud and Abuse Act (CFAA) and other statutes to push an idealist into the abyss.

JSTOR (Journal Storage) is one of the world's largest digital archives of academic journals, hosting over 12 million scholarly articles. The academic publishing industry has a long-criticized structural problem: researchers use public funds (government grants, university budgets) to conduct research, write papers for journals and perform peer review for free, yet the final publications are locked behind paywalls costing thousands to tens of thousands of dollars per year by commercial publishing giants like Elsevier and Springer Nature. Researchers and the public in many developing countries cannot access knowledge produced with taxpayer money. Swartz's actions were a direct challenge to this model.

Ironically, what Swartz downloaded were academic papers—content that should inherently serve the free flow of knowledge. He didn't sell the data for profit, didn't commercialize it, and was simply trying to break down information barriers. Yet the judicial machinery launched a disproportionate pursuit with a "make an example of him" posture.

Meta's "Compliance": Mass-Scraping Copyrighted Content to Train AI

In sharp contrast, Meta has been accused of obtaining tens of thousands of copyrighted books from LibGen and other piracy sources to train its Llama series of large language models. Court documents even show that internal company discussions weighed the legal risks of using pirated data, but ultimately decided to "proceed anyway."

Llama (Large Language Model Meta AI) is Meta's series of open-source large language models released progressively since 2023, competing in performance with closed-source models like GPT-4. In multiple class-action lawsuits and court document disclosures, plaintiffs allege that Meta used data from LibGen (Library Genesis, the world's largest pirated academic book and paper database, hosting over 3 million books) and Z-Library, among other sources. Internal communications show that Meta employees explicitly discussed the legal risks of using such data—some suggested using "clean" datasets, but due to competitive pressure and data scale requirements, the company ultimately decided to continue using the controversial data. Notably, OpenAI, Google, and other AI companies face similar allegations; this is not unique to Meta but a systemic practice across the entire generative AI industry.

However, Meta faces only civil lawsuits and has repeatedly prevailed in multiple rulings—courts tend to find that AI training constitutes "fair use." No executives have been charged, no one faces prison time, and the entire affair is merely a footnote in the operations of a company worth over a trillion dollars.

Structural Problems Behind the Double Standard

Power and Scale Determine How Law Is Applied

A recurring theme in the Hacker News discussion is that law enforcement often depends on "who you are" rather than "what you did." Swartz was an independent individual without a corporate legal army, without lobbying funds, without political influence; Meta has top-tier legal teams, deep political-business connections, and virtually unlimited litigation resources.

When the same conduct occurs at the hands of an individual versus a giant, the former faces the crushing weight of criminal felony charges while the latter bears only manageable business costs. This asymmetry exposes deep flaws in how the judicial system handles power differentials.

The CFAA's Vagueness and Prosecutorial Abuse

The core legal instrument in the Swartz case—the Computer Fraud and Abuse Act—has long been criticized for its vague language and susceptibility to arbitrary prosecutorial interpretation. The boundaries of "unauthorized access" are extremely ambiguous, enabling prosecutors to bring felony charges against virtually any data access activity.

The Computer Fraud and Abuse Act was originally passed in 1986, inspired in part by the Hollywood film WarGames (1983), in which a teenage hacker breaks into a military system. The act was initially designed to combat intrusions into government and financial institution computer systems, but after multiple amendments, its scope was vastly expanded. The phrase "exceeds authorized access" is extremely vague—theoretically, even violating a website's terms of service could be interpreted as a federal felony. Organizations like the Electronic Frontier Foundation (EFF) have long called for the act's reform. In 2021, the U.S. Supreme Court in Van Buren v. United States somewhat narrowed the CFAA's scope, ruling that merely accessing information one is authorized to access but using it for improper purposes does not violate the CFAA. However, the act's application to data scraping remains contentious.

When corporate-scale mass data scraping is involved, the issue is deftly reframed as a "fair use" debate within copyright law, sliding from the criminal domain into the civil one. The same act of "scraping" receives entirely different legal treatment.

Data Ethics and Copyright Dilemmas in the AI Era

The Boundaries of "Fair Use" Are Being Reshaped by AI Training

The current legal disputes over AI training data are essentially establishing the rules for how the entire AI industry acquires data. If courts continue to find that mass-scraping copyrighted content for AI training constitutes "fair use," why couldn't the same logic have applied to Aaron Swartz's actions?

Section 107 of U.S. copyright law establishes "fair use" as a case-law framework in which courts must weigh four factors: (1) the purpose and character of the use, including whether it is commercial or transformative in nature; (2) the nature of the copyrighted work; (3) the amount and substantiality of the portion used relative to the whole; and (4) the effect of the use on the work's potential market. In the context of AI training, multiple courts have leaned toward finding that "digesting" text into model weights is a highly transformative use—the output is entirely different in form from the input, and the model doesn't store and reproduce the original text but learns statistical patterns of language. However, this determination remains highly contested, especially when models can near-verbatim reproduce training data given certain prompts, making the "transformative" argument appear fragile. Different courts have reached inconsistent rulings, and the matter may ultimately require resolution by the U.S. Supreme Court or Congressional legislation.

This question touches on a fundamental contradiction in the intellectual property system: are we protecting creators' rights, or maintaining incumbents' data monopolies? When Meta can "digest" entire pirated libraries to train commercial models while a young person trying to liberate academic papers paid with his life, the disparity is impossible to ignore.

Implications for Open Source and the Open Knowledge Movement

The Open Access ideals that Swartz championed during his lifetime feel both more important and more fragile in the age of large AI models. On one hand, AI development genuinely requires vast amounts of data to thrive; on the other, the means of data acquisition, the distribution of rights, and the assignment of responsibility lack fair and consistent rules.

The Open Access movement made significant progress after Swartz's death. In 2018, a coalition of European research funding agencies launched Plan S, requiring all research they fund to be published in open access. In 2022, the White House Office of Science and Technology Policy (OSTP) issued a directive requiring all federally funded research papers to be freely available to the public immediately upon publication, eliminating the previously allowed 12-month embargo period. The flourishing of preprint servers like arXiv and bioRxiv is also transforming the landscape of academic communication. However, the relationship between open access and AI training data rights remains complex—even if papers are freely readable, whether they can be used for commercial AI training is an unresolved legal and ethical question. The goals Swartz fought for are being partially realized, but new challenges have already surpassed the problem domain he originally faced.

This discussion reminds us that technological progress should not serve as a fig leaf for institutional injustice. If the law gives giants a pass while throwing the book at individual idealists, then the so-called "innovation-friendly" environment is essentially just power crowning itself.

Conclusion: Justice in Data Scraping Should Not Have a Double Standard

Aaron Swartz's tragedy and Meta's composure form the most glaring contrast of the digital age. This is not merely a technical debate about the legality of data scraping—it is a profound questioning of legal justice, checks on power, and the ethics of knowledge.

As we celebrate AI's rapid advancement, perhaps we should remember: a system of rules that is harsh on the weak and lenient with the powerful will ultimately erode society's fundamental trust in the rule of law. The question Swartz posed with his life still has not received a just answer.