Last week, the U.S. District Court for the Northern District of California stunned the tech and legal communities by issuing two contradictory rulings on copyright cases involving artificial intelligence (AI) training within just 48 hours. The two cases, one targeting Anthropic and the other Meta, both saw judges ruling that AI training constitutes "fair use," but there were significant differences in their legal interpretations, highlighting the awkward position of current copyright laws when facing emerging technologies.

First, Judge William Alsup in the Anthropic case ruled that the company's AI training constituted transformative use, meaning their actions met the criteria for fair use under copyright law. Judge Alsup argued that the way AI models are trained is similar to human learning, and thus this type of use is acceptable. However, just 48 hours later, another judge, Vincent Chhabria, in the Meta case took a different view, emphasizing the fundamental differences between human learning and AI training, noting that Meta could not be simply compared to human learning processes.

In both rulings, the judges acknowledged that the works used in AI training have significant creative expressive value, rather than merely utilizing their functional elements. This consensus somewhat supports the stance of AI companies, but both judges' assessments of market harm were remarkably simplified. They barely explored potential market losses caused by AI training and even suggested that the existing market did not necessarily need protection.

Notably, the narrowness of these rulings should not be overlooked. Both judges clearly stated that their decisions were limited to specific circumstances, and future outcomes could vary greatly if new evidence or different usage scenarios arose. All of this indicates that although the current rulings provide some protection for AI companies, future legal challenges will continue, and the scope and rules of copyright law need to be constantly adjusted to keep up with the rapid development of technology.

This debate over AI and copyright law is like building a "torn" bridge between law and technology, challenging traditional legal frameworks and forcing us to think about how future copyright protection can adapt to emerging technologies.