Anthropic has asked a US federal court to declare its use of copyrighted song lyrics in training its AI chatbot Claude as fair use, sparking debate over AI’s impact on music copyrights. Meanwhile, India advances proposals for mandatory licensing to regulate AI training on protected content.
- US court weighing fair use for AI training on copyrighted lyrics
- India proposes mandatory blanket licensing for AI training works
- Major music publishers challenge AI-generated lyrics’ market impact
What happened
On October 21, Anthropic requested a US federal judge to rule that using copyrighted song lyrics to train its AI chatbot Claude falls under the fair use doctrine. The case, brought by leading music publishers including Universal Music Group, alleges unauthorized use of lyrics from artists like Beyoncé and the Rolling Stones during Claude’s training.
The lawsuit has prompted extensive legal arguments over the application of copyright law to AI training. In parallel, India’s Department for Promotion of Industry and Internal Trade proposed a mandatory licensing framework to regulate use of copyrighted materials in AI, requiring royalty payments determined by a government committee. However, this framework is not yet law.
Why it matters
This US case is crucial as it could set or clarify legal precedents for how copyrighted content can be used to develop AI systems, balancing innovation with the rights of content creators. The music publishers argue that AI-generated content threatens the economic interests of songwriters and undermines copyright protections.
India’s move toward mandatory licensing highlights the global concern over AI’s impact on intellectual property. Instead of relying solely on fair use or fair dealing exceptions, India aims to create a uniform compulsory licensing mechanism to ensure rights holders are compensated as AI technologies become commercially widespread.
What to watch next
The US federal judge’s forthcoming ruling on fair use could influence similar cases pending against Anthropic and other AI developers. Divergent earlier rulings in California’s Northern District demonstrate the complexity and uncertainty of applying copyright law to AI training datasets.
In India, stakeholders await potential legislation or regulation following public consultation on the proposed blanket licensing scheme. Meanwhile, ongoing litigation in Indian courts concerning AI training on copyrighted works, such as the ANI v OpenAI case, is likely to shape the country’s approach to AI and IP protection.