Japan Proposes New IP Safeguards for Generative AI Training

Japan's Cabinet Office has put forward a draft of new principles concerning the protection of intellectual property (IP) when training generative artificial intelligence (AI). The proposal seeks to increase transparency around how companies collect and utilize data for their AI models. Key elements include requirements to respect access restrictions, such as paywalls, and to strive to avoid crawling so-called pirate sites.

The draft operates on a "comply or explain" basis, where generative AI businesses must either adhere to the principles or publicly explain why they deviate. The objective is to balance the development of generative AI with the protection of rights holders' interests. These new guidelines would apply to both Japanese and foreign companies offering their services in Japan.

The proposal emphasizes businesses' responsibility to safeguard intellectual property rights and mandates annual reviews and public disclosures of these principles. Companies are also expected to maintain logs of training data usage and implement technical measures, such as digital watermarking, to prevent copyright-infringing outputs.

Japan's move comes amidst ongoing international discussions on the legal status of AI training data. The country's existing copyright law permits, under specific conditions, the use of copyrighted material for data analysis, including AI training, without explicit permission. The new proposal does not amend the law itself but introduces complementary governance practices and transparency requirements.