Japan Releases Draft AI Guidelines Urging Developers to Disclose Training Data
Japan's Cabinet Office has released a draft guideline urging generative AI developers to disclose their training data, collection methods, and whether copyrighted materials are used. The guidelines also advise developers to avoid sourcing data from pirated websites. This move represents a significant policy shift aimed at balancing rapid AI innovation with the protection of creators' intellectual property rights. It addresses growing concerns over copyright infringement in generative AI models, particularly regarding creative works like manga, music, and movies. Although the draft guidelines are currently non-binding, they encourage developers to establish internal notification systems to alert users when training datasets contain copyrighted works similar to the generated output.
## BACKGROUND
Generative AI models require massive datasets, often scraped from the internet, to train and generate text, images, or music. This practice has sparked global legal and ethical debates, as creators argue that their copyrighted works are being used without consent or compensation.