Recently, a music publisher filed a preliminary injunction against Anthropic on the grounds that Anthropic used copyrighted music content in the training of its large language model (LLM). The incident highlights the controversial issue of copyright in training data for artificial intelligence models. Anthropic countered that using copyrighted content for training is reasonable and rejected the injunction request, arguing that obtaining all licenses would stifle the development of current general-purpose AI tools.
The article focuses on:
The music publisher filed a preliminary injunction against Anthropic, and Anthropic responded by pointing out that the use of copyrighted content as LLM training data was reasonable and rejecting the injunction request. The company believes that training systems on copyrighted content to obtain permissions will mean the end of today's general-purpose AI tools.
The core of this dispute lies in the definition and reasonable use of copyright of artificial intelligence model training data. Anthropic’s response reflects a common dilemma facing the AI industry: the balance between access to training data and copyright protection. Future developments may rely on clearer legal frameworks and technological solutions.