On Wednesday, in the AI copyright litigation Kadrey, et al. v. Meta Platforms Inc., the judge ruled in favor of Meta. U.S. District Judge Vince Chhabria conveyed hesitance, characterizing his verdict as “in significant tension with reality.”

Thirteen writers, including Sarah Silverman, Ta-Nehisi Coates, and Junot Diaz, filed a lawsuit against Meta for employing their books without authorization to develop its Llama AI models.

The facts of the case were remarkable. Meta not only utilized unauthorized copies of the authors’ creations but internal communications indicated that Meta staff had legal and ethical reservations about this. Additional messages hinted at attempts to eliminate evidence of piracy, looking for phrases like “stolen” and “pirated.”

Chhabria’s decision complicates the legal matter rather than resolving the copyright dispute regarding AI training.

A day prior, a judge in a related case ruled in favor of Anthropic. U.S. District Judge William Alsup determined in Bartz v. Anthropic that utilizing pirated books in shadow libraries Books3 and LibGen constituted fair use.

Robert Brauneis, a professor specializing in intellectual property law, pointed out that Judges Alsup and Chhabria employed differing reasoning. Both cases relied on the fair use doctrine, particularly the fourth factor — potential harm to the market.

Brauneis elaborated that Judge Alsup maintained a narrow perspective: if AI-generated output does not infringe upon a work utilized in training, any sales loss cannot be deemed market harm. Judge Chhabria disagreed, asserting that market dilution harm should be considered and could surpass transformativeness.

Both judges supported fair use, yet their varying rationales produced a complex legal setting.

The Kadrey plaintiffs were unable to counter Meta’s fair use assertion. Kevin Madigan from the Copyright Alliance observed that the plaintiffs’ lawyer did not offer counterarguments.

The litigation centered on whether the usage was transformative and if it harmed the market for the copyrighted material. Chhabria sided with Meta on the transformative aspect, indicating that Meta’s usage had a “further purpose” and “different character.”

The fourth factor, market harms, posed challenges for the plaintiffs. Chhabria scrutinized their market harm evidence, presenting a hypothetical regarding future artists like Taylor Swift.

Chhabria advised the plaintiffs’ counsel that they would face loss if they couldn’t demonstrate substantial market impact.

Ultimately, Chhabria ruled in favor of Meta, asserting that the plaintiffs failed to provide evidence of market harm.

A Meta representative stated, “We appreciate today’s ruling. Open-source AI models are driving transformative innovations, and fair use is essential for developing this technology.”

Chhabria’s decision was primarily focused on the deficiencies in the plaintiffs’ arguments rather than the fair use defense itself. He indicated the judgment only impacted these thirteen authors, not others whose works Meta had utilized.

The verdict leaves an opening for other creators to initiate similar legal actions against Meta and additional AI firms. Chhabria hinted that copying works to train AI models without authorization might be unlawful.

This ruling carries symbolic significance for artists. Justin Hughes, a law professor, cautioned that designating training AI on pirated datasets as fair use would be an unfortunate outcome for creative professionals.

Kadrey v. Meta is one of numerous copyright lawsuits against AI companies. AI is already influencing creative sectors, with generative AI facilitating the automation of text, images, video, and audio creation, potentially displacing creative jobs.

A report from the Animation Guild indicated that AI could consolidate or displace current job titles, potentially resulting in a loss of 203,800 jobs by 2026.

Numerous artists perceive AI tools like Llama as a menace, trained on human expression they are accused of supplanting.

The American Association of Publishers contended that Meta’s actions would yield dire consequences for authors and publishers.

While Meta emerged victorious, Madigan described Chhabria’s ruling as a “mixed bag.” The plaintiffs may seek to appeal, as their spokesperson expressed disagreement with the judgment.

Kadrey v. Meta and Bartz v. Anthropic concentrate on pirated books as AI training resources, whereas other cases against OpenAI and Microsoft emphasize AI outputs.

In cases with evidence of infringing output, Kadrey v. Meta is not applicable. These cases stress demonstrating side-by-side verbatim reproductions instead of abstract debates regarding transformative use.