Decoding the ‘Fair Use’ Doctrine in the AI Era

Last week, Universal Music and other major record labels sued Anthropic, an Amazon and Google backed AI platform. Plaintiff’s complaint alleged that Anthropic’s chatbot, known as Claud 2, illegally distributed copyrighted song lyrics. The past couple of months have seen a slew of lawsuits against tech companies who produce Large Langue Model (LLM) based chatbots.

According to Reuters, the lawsuit accused Anthropic of infringing on publishers’ copyrights by copying their lyrics without permission. This is done by scraping ‘massive amounts of text’ from the internet, in order to train Claude to respond to human prompts.

Plaintiff’s complaint stated that when Claud is prompted to create a new song or poem, it pulls lyrics directly from copyrighted songs instead of creating an original narrative. When this happens, the new creation does not include the corresponding songwriter or song title –critical copyright information. The music labels argued that Anthropic is different from typical lyric-sharing websites because Anthropic’s Claud does not credit the original work, thereby evading established standards in copyright law.  

Fair use or copyright infringement?

According to the U.S. Copyright Office, the fair use doctrine is a vital aspect of American copyright law. Fair use, like free speech, is vague in meaning. Fair use is a judge created doctrine codified in the 1976 Copy Right Act. Essentially, fair use is a legal defense against a claim of copyright infringement. It is “any copying of copyrighted material done for a limited and ‘transformative’ purpose, such as to comment upon, criticize, or parody a copyrighted work.” If something qualifies as fair use, it can’t be considered an infringement.

Litigation centers on the meaning of “transformative,” as millions of dollars in legal fees are spent each year attempting to define what qualifies as fair use. Judges wanted the fair use exception to have a broad definition, leaving it open to many interpretations under the law.

We are living in an era where AI is changing the landscape of artistic industries. Many have argued that copyright law and the fair use doctrine must be redefined in order to regulate AI. For example, a portion of the current strike between the Writers Guild of America and Hollywood studios centers around the use of AI. The Writers Guild wants to protect writers from copyright infringement, while the studios want the right to use writers’ work to train AI models.

Another series of high profile lawsuits concern OpenAI (which owns ChatGPT) clashing with writers and artists on what content should be allowed into the knowledge base of generative artificial intelligence. Under current Copyright law, limited portions of someone’s work can be be used without a license or compensation. But what happens when generative AI and language-based model systems index entire bodies of work within their knowledge base? Many argue it is not fair use, and as such, a copyright infringement.

The latest lawsuit from Universal Music group highlights how AI is changing the landscape of the music industry. At the heart of the matter is whether artistic works can be legally used to train AI software without breaching copyright and intellectual property laws. Record labels, authors, and actors are battling tech companies in the courts to determine whether tech companies should compensate folks when they use their artistic works to train AI chatbots.

Finding a way forward

The outcome of these lawsuits will lay the groundwork for new laws governing how AI engines are trained, and whether tech companies must compensate artists if a chatbots’ output content reproduces part or all of a copyrighted work.

AI companies argue that the training of their software is fair use, simply redefining the human process of reading existing information and generating new ideas. Though compelling, this argument presupposes that AI is equivalent to human intelligence. Overlooking a key difference, which is that generative AI’s ability to take in and learn information is basically limitless, whereas a human’s is not.

The music business has become an early legal testing ground for generative AI, and perhaps this industry will be the first to find a commercial path forward. Because AI large-language models are relatively new, existing legal precedent on the matter is scarce. This is makes it difficult to predict how the courts will rule. But one thing is for sure: AI will redefine copyright law and how the fair use doctrine is applied to intellectual property.

Copyright and Intellectual property are complex areas of law. They aim to strike a balance between creative freedom and compensating those who have created original work. In the age of the internet and progressing AI technology, the line between original content and creative license is blurry. The courts and legislators must differentiate between human learning and AI capacity in order to strike a balance between fair use and copyright infringement.

Interested in copy right law? Want to learn more about the fair use doctrine? Head on over to and browse through hundreds of dockets filled with similar cases.