TOKYO: Japan’s Cultural Affairs Council subcommittee Thursday broadly accepted government-presented draft ideas regarding copyright infringements associated with generative artificial intelligence.

The Cultural Affairs Agency drafted the ideas based on public opinions in a bid to curb unauthorized training of generative AI with copyrighted materials.

The agency will promote the ideas after they are reported at a council meeting on the matter next month.

The copyright law stipulates that if data is learned by AI at the development stage, copyrighted works can be used without permission, in principle.

Meanwhile, permission is required when the interests of copyright holders are feared to be infringed on, while it is not required for minor use of copyrighted things.

According to the draft ideas, permission will be made mandatory for training AI to output creative expressions of a specific creator.

The combined use of search engines and AI to process and summarize the latest online information will no longer fall into the “minor use” category if it goes beyond a certain level.

Citing database sales by European and U.S. newspaper publishers as examples of copyright holders’ interests being unreasonably harmed, the agency said permission will be required for having AI learn databases expected to be commercialized by bypassing technical measures to prevent such learning.

Moreover, it suggested the possibility of AI developers being held accountable for copyright infringements if they collect learning materials while knowing they are pirate versions.

JIJI Press