朝活ミニテスト 第38回

Who Owns AI Art?

Generative artificial intelligence can now produce images, music, and text that closely resemble human work. This ability has raised difficult legal questions, the first of which concerns input. AI systems are trained on enormous collections of existing material, much of it protected by copyright and gathered from the internet without the creators' permission. Technology firms argue that this is permitted under principles such as fair use in the United States, because the systems learn general patterns rather than storing and copying specific works. In their view, training an AI resembles the way human artists develop by studying the works of others.

Many creators strongly disagree. They note that AI systems can sometimes reproduce passages, melodies, or visual styles almost exactly, and that AI-generated works may compete directly with the human artists whose creations were used to train them. Several lawsuits brought by authors, artists, and news organizations are now working their way through the courts, and their outcomes may shape the industry for years. Some companies have responded by signing licensing agreements that pay publishers or image libraries for access to their material.

The second question concerns output. In many countries, copyright protection has traditionally required human authorship. The United States Copyright Office, for example, has refused to register works created entirely by AI, although it may protect human contributions such as selecting and arranging generated images. Critics of this position argue that the line between human and machine creativity is becoming increasingly blurred as people use AI as a tool at many stages of their work. As a result, lawmakers face growing pressure to clarify how much human involvement is needed for a work to be protected.