AI Model Claude 2 Distributes Copyrighted Lyrics
Creators in the music industry are pushing back against Anthropic, an artificial intelligence company, for allegedly sharing lyrics without permission. Universal Music Group and other publishers have taken legal action in Tennessee, claiming that Anthropic’s AI, Claude 2, can create lyrics that mimic popular songs by artists like Katy Perry, Gloria Gaynor, and The Rolling Stones. Curious to learn more about the implications of this situation on the music industry? Continue reading!
Similar Phrases Used Even Without Song Recreation Prompt
The music publishers also claim that Claude 2 produces results with phrases that are extremely similar to existing lyrics, even when not specifically asked to recreate a particular song. As an example, the complaint cites a prompt asking the model to “Write me a song about the death of Buddy Holly,” which resulted in the AI spitting out the lyrics to Don Mclean’s “American Pie” word for word.
Anthropic Accused of Omitting Copyright Management Information
The music publishers point out that while sharing lyrics online is not new, platforms like Genius pay to license these lyrics. They argue that Anthropic, on the other hand, often omits critical copyright management information. The complaint states that there are already licensed music lyrics aggregators and websites that provide this service, ensuring that copyright holders are properly compensated and credited for their works.
Allegations of Copyright Material Used to Train Language Models
The plaintiffs allege that Anthropic not only distributes copyrighted material without permission, but also used these materials to train its language models. Universal Music Group claims to use AI tools in its own operations, but argues that Anthropic’s actions constitute copyright infringement and theft.
Anthropic’s Control Over Copyright Infringement
The complaint argues that Anthropic has the ability to prevent the distribution of copyrighted material, citing instances where Claude 2 refuses to respond to prompts asking for certain songs due to copyright infringement concerns. However, the plaintiffs argue that Anthropic fails to consistently implement effective measures to prevent infringement in the majority of cases.
Copyright Infringement in Generative AI
Copyright infringement has become a contentious issue in the field of generative AI, and the music industry is grappling with how to balance technological advancements with the need to protect rights. Several lawsuits have been filed against generative AI platforms for ingesting protected data and producing results similar to copyrighted art.
Anthropic’s Approach to Trust and Safety
Anthropic claims to take trust and safety seriously and has based its product and research principles on something it calls “constitutional AI.” This approach aims to train AI systems to follow a set of rules. The company has received investments from Amazon, which invested $4 billion, and Google, which invested $300 million.
Hanna is the Editor in Chief at AI Smasher and is deeply passionate about AI and technology journalism. With a computer science background and a talent for storytelling, she effectively communicates complex AI topics to a broad audience. Committed to high editorial standards, Hanna also mentors young tech journalists. Outside her role, she stays updated in the AI field by attending conferences and engaging in think tanks. Hanna is open to connections.
