Sony AI Unveils "Protective AI" to Block Ghibli-Style Copies
Sony’s new research tool trains on copyrighted works to prevent imitation in generative media while tracking creator attribution.
Summary
- Sony AI is developing a “Protective AI” tool designed to stop generative models from creating copyright-infringing content
- The system is trained on copyrighted material, like Studio Ghibli films, so it learns exactly what to avoid generating
- Sony hopes the tool will also track attribution to ensure original creators are fairly compensated for their work
Sony‘s research and development division, Sony AI, is currently building a new technology dubbed “Protective AI” aimed at curbing copyright infringement in AI-generated media. First reported by Nikkei, the tool is specifically designed to stop AI models from producing unauthorized imitations of existing intellectual property, such as videos or music that mimic the distinctive art style of Studio Ghibli. To achieve this, Sony is taking the counterintuitive approach of deliberately feeding copyrighted data directly into the system, training the model to recognize and strictly avoid those specific styles even when users attempt to bypass restrictions with indirect prompts.
Beyond simply blocking plagiarized content, Protective AI is also being developed to address ongoing concerns regarding artist compensation. Sony aims to use the underlying tracking technology to identify when a creator’s original work has contributed to an AI-generated output. While the tool is still in its research and development phase with no set timeline for internal deployment, the initiative could eventually pave the way for a fair compensation framework across Sony’s extensive portfolio of games, music, movies, and anime.
