Creating and sharing knowledge are defining traits of humankind, yet copyright law has grown so restrictive that it can require acts of civil disobedience to ensure that students and scholars have the books they need and to preserve swaths of culture from being lost forever.

Reputable research generally follows a familiar pattern: Scientific articles are written by scholars based on their research—often with public funding. Those articles are then peer-reviewed by other scholars in their fields and revisions are made according to those comments. Afterwards, most large publishers expect to be given the copyright on the article as a condition of packaging it up and selling it back to the institutions that employ the academics who did the research and to the public at large. Because research is valuable and because copyright is a monopoly on disseminating the articles in question, these publishers can charge exorbitant fees that place a strain even on wealthy universities and are simply out of reach for the general public or universities with limited budgets, such as those in the global south. The result is a global human rights problem.

This model is broken, yet science goes on thanks to widespread civil disobedience of the copyright regime that locks up the knowledge created by researchers. Some turn to social media to ask that a colleague with access share articles they need (despite copyright’s prohibitions on sharing). Certainly, at least some such sharing is protected fair use, but scholars should not have to seek a legal opinion or risk legal threats from publishers to share the collective knowledge they generate.

Even more useful, though on shakier legal ground, are so-called “shadow archives” and aggregators such as SciHub, Library Genesis (LibGen), Z-Library, or Anna’s Archive. These are the culmination of efforts from volunteers dedicated to defending science.

SciHub alone handles tens of millions of requests for scientific articles each year and remains operational despite adverse court rulings thanks both to being based in Russia, and to the community of academics who see it as an ethical response to the high access barriers that publishers impose and provide it their log-on credentials so it can retrieve requested articles. SciHub and LibGen are continuations of samizdat, the Soviet-era practice of disobeying state censorship in the interests of learning and free speech.

Unless publishing gatekeepers adopt drastically more equitable practices and become partners in disseminating knowledge, they will continue to lose ground to open access alternatives, legal or otherwise.

EFF is proud to celebrate Open Access Week.

The White House’s recently-unveiled “AI Action Plan” wages war on so-called “woke AI”—including large language models (LLMs) that provide information inconsistent with the administration’s views on climate change, gender, and other issues. It also targets measures designed to mitigate the generation of racial and gender biased content and even hate speech. The reproduction of this bias is a pernicious problem that AI developers have struggled to solve for over a decade.

A new executive order called “Preventing Woke AI in the Federal Government,” released alongside the AI Action Plan, seeks to strong-arm AI companies into modifying their models to conform with the Trump Administration’s ideological agenda.

The executive order requires AI companies that receive federal contracts to prove that their LLMs are free from purported “ideological biases” like “diversity, equity, and inclusion.” This heavy-handed censorship will not make models more accurate or “trustworthy,” as the Trump Administration claims, but is a blatant attempt to censor the development of LLMs and restrict them as a tool of expression and information access. While the First Amendment permits the government to choose to purchase only services that reflect government viewpoints, the government may not use that power to influence what services and information are available to the public. Lucrative government contracts can push commercial companies to implement features (or biases) that they wouldn't otherwise, and those often roll down to the user. Doing so would impact the 60 percent of Americans who get information from LLMs, and it would force developers to roll back efforts to reduce biases—making the models much less accurate, and far more likely to cause harm, especially in the hands of the government. 

Less Accuracy, More Bias and Discrimination

It’s no secret that AI models—including gen AI—tend to discriminate against racial and gender minorities. AI models use machine learning to identify and reproduce patterns in data that they are “trained” on. If the training data reflects biases against racial, ethnic, and gender minorities—which it often does—then the AI model will “learn” to discriminate against those groups. In other words, garbage in, garbage out. Models also often reflect the biases of the people who train, test, and evaluate them. 

This is true across different types of AI. For example, “predictive policing” tools trained on arrest data that reflects overpolicing of black neighborhoods frequently recommend heightened levels of policing in those neighborhoods, often based on inaccurate predictions that crime will occur there. Generative AI models are also implicated. LLMs already recommend more criminal convictions, harsher sentences, and less prestigious jobs for people of color. Despite that people of color account for less than half of the U.S. prison population, 80 percent of Stable Diffusion's AI-generated images of inmates have darker skin. Over 90 percent of AI-generated images of judges were men; in real life, 34 percent of judges are women. 

These models aren’t just biased—they’re fundamentally incorrect. Race and gender aren’t objective criteria for deciding who gets hired or convicted of a crime. Those discriminatory decisions reflected trends in the training data that could be caused by bias or chance—not some “objective” reality. Setting fairness aside, biased models are just worse models: they make more mistakes, more often. Efforts to reduce bias-induced errors will ultimately make models more accurate, not less. 

Biased LLMs Cause Serious Harm—Especially in the Hands of the Government

But inaccuracy is far from the only problem. When government agencies start using biased AI to make decisions, real people suffer. Government officials routinely make decisions that impact people’s personal freedom and access to financial resources, healthcare, housing, and more. The White House’s AI Action Plan calls for a massive increase in agencies’ use of LLMs and other AI—while all but requiring the use of biased models that automate systemic, historical injustice. Using AI simply to entrench the way things have always been done squanders the promise of this new technology.

We need strong safeguards to prevent government agencies from procuring biased, harmful AI tools. In a series of executive orders, as well as his AI Action Plan, the Trump Administration has rolled back the already-feeble Biden-era AI safeguards. This makes AI-enabled civil rights abuses far more likely, putting everyone’s rights at risk. 

And the Administration could easily exploit the new rules to pressure companies to make publicly available models worse, too. Corporations like healthcare companies and landlords increasingly use AI to make high-impact decisions about people, so more biased commercial models would also cause harm. 

We have argued against using machine learning to make predictive policing decisions or other punitive judgments for just these reasons, and will continue to protect your right not to be subject to biased government determinations influenced by machine learning.

When a technology can have a conversation with you, it’s natural to anthropomorphize that technology—to see it as a person. It’s tempting to see a chatbot as a thinking, speaking robot, but this gives the technology too much credit. This can also lead people—including judges in cases about AI chatbots—to overlook the human expressive choices connected to the words that chatbots produce. If chatbot outputs had no First Amendment protections, the government could potentially ban chatbots that criticize the administration or reflect viewpoints the administration disagrees with.

In fact, the output of chatbots can reflect not only the expressive choices of their creators and users, but also implicates users’ right to receive information. That’s why EFF and the Center for Democracy and Technology (CDT) have filed an amicus brief in Garcia v. Character Technologies explaining how large language models work and the various kinds of protected speech at stake.

Among the questions in this case is the extent to which free speech protections extend to the creation, dissemination, and receipt of chatbot outputs. Our brief explains how the expressive choices of a chatbot developer can shape its output, such as during reinforcement learning, when humans are instructed to give positive feedback to responses that align with the scientific consensus around climate change and negative feedback for denying it (or vice versa). This chain of human expressive decisions extends from early stages of selecting training data to crafting a system prompt. A user’s instructions are also reflected in chatbot output. Far from being the speech of a robot, chatbot output often reflects human expression that is entitled to First Amendment protection.

In addition, the right to receive speech in itself is protected—even when the speaker would have no independent right to say it. Users have a right to access the information chatbots provide.

None of this is to suggest that chatbots cannot be regulated or that the harms they cause cannot be addressed. The First Amendment simply requires that those regulations be appropriately tailored to the harm to avoid unduly burdening the right to express oneself through the medium of a chatbot, or to receive the information it provides.

We hope that our brief will be helpful to the court as the case progresses, as the judge decided not to send the question up on appeal at this time.

Read our brief below.