Two major AI companies Meta and Anthropic recently secured wins in US copyright lawsuits regarding the use of copyrighted content to train their large language models (LLMs). While these summary judgments were hailed by some as landmark decisions, a closer look shows that the rulings are narrow and far from conclusive.
1. Courts Lean Toward Fair Use For Now
In Anthropic’s case, US District Judge William Alsup ruled that using a database of scanned, purchased books to train the Claude model qualified as “transformative” and therefore met the fair use standard under US copyright law. Similarly, Meta’s Llama model training was also found to fall within the bounds of fair use by Judge Vince Chhabria. However, both judges emphasized that their rulings were specific to the facts of each case.
2. The Decisions Are Not Broad Precedents
Despite media portrayal, these are not sweeping victories that settle the debate over AI and copyright. The judgments address only whether the training processes used in these specific instances were fair use. They do not offer blanket protection for all AI training using copyrighted material.
3. Legal Questions Remain Unanswered
These cases did not resolve when, during the AI development process, a copyright violation may occur—whether during training, inference, or model output. Nor do they answer whether creators deserve compensation for their works being used without permission. These core issues remain largely untested in court.
4. Copyright Law Is Struggling to Catch Up
The rulings expose how traditional copyright law is still grappling with modern AI practices. Legal frameworks are not yet fully equipped to deal with the complexities of LLMs trained on massive datasets that include copyrighted content.
5. More Litigation Is Coming
These cases are just the beginning. Given the billions invested in AI and the potential value of copyrighted content, more lawsuits are expected. Courts will likely be asked to decide finer details, including the rights of authors and the responsibilities of tech companies.
In conclusion, Meta and Anthropic’s wins offer temporary relief but leave the broader legal landscape of AI and copyright murky and unsettled.
