Sony develops Protective AI to stop AI from stealing Studio Ghibli
Sony AI division trains a model on the studio's data to block imitations and calculate compensation for creators
Image from the original Anitsu edition. Sony AI's research team announced the development of Protective AI (PA), a technology designed to prevent copyright infringement in videos and music generated by artificial intelligence. The system aims to create a mechanism so that creators and rights holders receive fair compensation when their work contributes to a generation.
A shield for Studio Ghibli
The tool was trained specifically on Studio Ghibli data to detect and reject imitations. According to Automaton, Protective AI is designed to prevent copying even through indirect prompts: the model receives the original material and is instructed to generate results that move away from that style.
Royalty collection
The project also contemplates attribution and compensation for rights holders. Sony AI had already mentioned similar efforts in music, citing the experimental use of unlearning algorithms to measure how much each training example contributed to the generated output. Protective AI would apply a similar approach to video.
The initiative remains in the research and development phase. Sony has not yet adopted the technology internally nor announced a public release date.
Originally published on Anitsu by Kim Seo-yeon. Text reviewed and cross-checked for DonOtaku on November 5, 2026; the original date is preserved.
0 comments
Sign in to leave a comment
Tell everyone what you think about this story.
No comments yet. Be the first.
Keep reading

Being a streamer is now the number one dream of Japanese children, but it disappears in high school
Staff Donotaku · 1 April 2026

Cosplay photos and AI: how to recognize and report impersonation
Staff Donotaku · 30 March 2026

One Piece challenges you to become the most muscular Luffy
Juan Castillo · 9 February 2025