EFF warns that narrowing fair use to stop AI training could hurt everyone

The scope of the copyright fight

Publishers and creators are currently suing large AI companies for using copyrighted works to train generative models. While these cases are often framed as a battle between authors and Big Tech, the Electronic Frontier Foundation (EFF) argues the stakes are much broader. They believe the outcome will affect billions of people who are not involved in the lawsuits.

The EFF recently filed an amicus brief in the case of In re OpenAI to explain these broader implications. They argue that the court's ruling on fair use—the legal doctrine allowing limited use of copyrighted material without permission—will have a ripple effect. If the law is narrowed to stop AI training, it could simultaneously impede work that benefits the public.

The organization emphasizes that while creators deserve support and tech companies must be held accountable, giving copyright holders more control over the web is not the solution. They suggest that the plaintiffs are seeking to rewrite copyright law in ways that favor only the most powerful media and tech companies. This shift would potentially harm researchers, activists, and people with disabilities who rely on open data.

How fair use protects the web

Fair use exists to stimulate creativity and innovation for the public good, rather than just protecting the first person to publish a work. The EFF argues that the plaintiffs' current arguments would thwart this basic purpose of copyright. They point out that AI models are not designed to be 'infringement machines' that churn out copies of their training data.

Instead, courts have already found that training models is a transformative use because it serves a fundamentally different purpose than the original work. If the law changes to require permission for every copy made without a license, it would threaten everyday internet use. The EFF notes that routine activities like opening emails, clicking links, or streaming music all require what they call 'intermediate copying.'

Without fair use protections, users might be compelled to track the source of every image or article they click on to ensure rights are cleared. The EFF warns that such a requirement would cast legal doubt on billions of copies made daily. They argue that fair use is what allows the general public to navigate the internet without constant legal hurdles.

Risks to research and journalism

The legal strategies used by plaintiffs could put beneficial research and reporting at risk. Journalists and nonprofit researchers often use techniques like 'crawling' or 'scraping' to gather information, which is virtually identical to the methods used by AI giants. These groups use these tools to investigate anticompetitive practices or research misinformation on platforms like X.

Because these researchers often lack the deep pockets of major tech companies, they would be disproportionately affected by new restrictions. If copying without permission is ruled out as fair use, the law would no longer protect these transformative uses. This could prevent watchdog groups from holding Big Tech accountable or uncovering important public interest stories.

Furthermore, the EFF argues that copyright should not prevent people from building on existing ideas. Under the plaintiffs' theory, it could become difficult for people to write books inspired by a favorite genre or for academics to publish literature reviews. These activities would be restricted simply because the new publications might compete with the older ones.

The danger of mandatory licensing

Some creators argue that AI developers should simply obtain a license for every work they use. However, the EFF argues that this would entrench the power of the largest companies and hurt smaller competitors. Because the copyright system is often inefficient, only a handful of tech giants would be able to afford the costs of such licenses.

The organization also suggests that creators might not actually benefit from these licensing arrangements. They point out that media giants have historically exploited underpaid artists for decades. There is no guarantee that these large companies will treat artists more fairly once they control the data that powers AI models.

The EFF compares trying to help creators through more copyright rights to giving a bullied kid more lunch money for a bully to take. They argue that the rules governing AI should serve the public interest rather than just the companies that can afford to litigate for decades. The goal should be to ensure that AI tools remain available for medicine, science, and accessibility.