Sony’s “Protective AI” Marks a Turning Point for Copyright Protection in the Age of Generative Media
- 13 hours ago
- 4 min read
A recent article mentioned a Sony AI tool in development that is worth taking a look at, see https://automaton-media.com/en/news/sony-has-developed-an-ai-tool-that-protects-creators-copyrights-by-prohibiting-ghibli-style-outputs-and-providing-compensation/.
In a world where generative AI tools can effortlessly imitate iconic artistic styles, the line between inspiration and infringement has become increasingly blurred. Sony’s latest research initiative, Protective AI (PA), aims to redraw that line with unprecedented precision — and in doing so, it may reshape the future relationship between creators and AI systems.
Disclaimer: This blog post is merely for informational purposes and is not intended to provide, nor should they be relied upon as, legal advice. Legal rules can change often. The information contained herein is accurate to the best of our knowledge as of the date of this post.
A New Approach to Copyright Protection
Sony AI, the company’s dedicated AI-focused R&D division, has developed Protective AI to tackle one of the biggest challenges in today’s generative landscape: preventing unauthorized imitation of copyrighted works. According to reporting from The Nikkei and Automaton West, the technology is engineered to block “Ghibli-style” outputs and prevent infringement even when users attempt indirect or subtle prompts.
Unlike typical generative models that create outputs from broad datasets, Protective AI takes a counterintuitive but powerful approach: it’s trained on the copyrighted material it aims to protect — including Studio Ghibli works — but instructed to deliberately generate content that diverges from those styles.
Supporting Creator Compensation
Beyond its protective function, Sony’s technology also moves toward solving another critical industry gap: fair attribution and compensation. Protective AI is designed to allow rights holders to be compensated when their creative works contribute to AI‑generated content, echoing Sony’s ongoing work with “unlearning algorithms” in music attribution.
This aligns with broader industry concerns as Ghibli-style AI art has proliferated across social platforms, often without the studio’s consent or artist attribution — a trend that has raised questions about derivative works and IP rights for generative models.
Still in R&D — But With Major Implications
Although Protective AI is still in the research and development stage and not yet used internally at Sony, its implications are significant. With Sony's deep portfolio across games, film, anime, and music, the tool could become a foundation for responsible AI integration across its creative ecosystem and potentially set an industry standard.
If successful, Protective AI could become a cornerstone in establishing ethical AI practices — preventing unauthorized stylistic imitation, ensuring transparency, and creating viable compensation models for creators whose works fuel innovation.
Latest Court Rulings
My previous blog post discussed the most recent court cases and their conclusions on AI and AI training in relation to copyrighted materials. They found that the use of copyrighted materials may be considered fair use if the materials are lawfully acquired, but not if they are improperly sourced or pirated. Additionally, the factor of market harm will still factor into any determination.
1️⃣ Training AI on lawfully acquired data can be fair use, and
2️⃣ Pirated or improperly sourced data is a legal liability.
In Bartz v. Anthropic, the court ruled that using lawfully obtained books to train an LLM was fair use because the process is transformative — but the same court allowed infringement claims to proceed where pirated materials were retained internally, even if not actively used for training.
Another case, Kadrey v. Meta, reaffirmed that while AI training can be transformative, market harm remains the most important factor, and fair use is not automatic.
In relation to Protective AI, it is likely that Sony will need to utilize copyrighted materials that they own or obtain lawfully to train the AI related to not copying the materials. It will be interesting to see how Sony expands the AI training to other materials and how the R&D phase progresses on this AI. There is certainly an appeal into the approach of training to avoid copyright infringement.
Why This Matters
As AI-generated content becomes more sophisticated, the ethics and legality of training data usage are increasingly under scrutiny. Sony’s Protective AI may represent a practical, scalable step toward responsibly balancing innovation with the rights of creators — many of whom are striving to protect the integrity and value of their artistic contributions.
This is more than just a tool — it’s a signal that the future of AI need not come at the expense of the creators who inspire it. However, it will still require understanding of already protected works and remain diligent on your own best practices on how to use AI, and how to make sure that creators are attributed and compensated appropriately.
Something else that will need to be considered in this ever evolving space may also be trademarks, as these may include registered trademark names, logos or phrases, and unregistered and protected by common law rights trademark names, logos or phrases. Right now many AI databases are built around already existing data, and this includes those protected by copyrights or trademarks or other protections, and the AI does not necessarily exclude protected results when it provides an answer. This requires human due diligence to check the results.
Keep Learning, Keep Creating, Keep Protecting
Your voice matters. Your creativity matters. And in an AI‑powered world, protecting your work matters more than ever.
If you want to go deeper, get clarity, or explore next steps, you can connect with me or follow aTMospheric IP @atmoip on:
If you have any questions for me, or if you would like to meet me, please book a free initial consultation appointment with me online at www.atmoip.com/book-online, contact me through my website at www.atmoip.com/contact or send me an email at info@atmoip.com.
Your creativity has value. Your rights protect it. AI doesn’t change that—knowledge does.
aTMospheric IP, LLC
HQ 371 NE GILMAN BLVD., STE 160
1ST FLOOR
ISSAQUAH, WA 98027
Comments