TLDR: A new US tech platform, Vermillio, has developed an AI tool capable of tracking copyrighted material within AI-generated content. The company uses ‘neural fingerprints’ to identify the approximate percentage of pre-existing copyrighted material used by AI models, aiming to create a more equitable ecosystem for content creators and AI companies.
In a significant development for the evolving landscape of artificial intelligence and intellectual property, the US tech platform Vermillio has introduced an innovative AI tool designed to track the use of copyrighted material within AI-generated content. This new technology addresses a critical concern for creators and copyright holders who are grappling with the unauthorized use of their work by generative AI models.
Vermillio’s approach involves creating a ‘neural fingerprint’ for various pieces of copyrighted work. This unique identifier allows the platform to trace, approximately, the percentage to which an AI-generated image or content has drawn on pre-existing copyrighted material. In research conducted for The Guardian, Vermillio demonstrated this capability by creating neural fingerprints for copyrighted works and then prompting AI systems to generate similar imagery.
Kathleen Grace, the chief strategy officer at Vermillio, emphasized the potential for a collaborative future. “We can all win if we just take a beat and figure out a way to share and track content,” Grace stated. She further elaborated on the benefits, suggesting that such a system would “incentivise copyright holders to release more data to AI companies and would give AI companies access to more interesting sets of data. Instead of giving all the money to five AI companies, there would be this amazing ecosystem.”
The introduction of Vermillio’s tool comes amidst a backdrop of increasing legal challenges and debates surrounding AI and copyright. Generative AI models, which power tools like OpenAI’s ChatGPT and Google’s Veo3, are trained on vast datasets often sourced from the open web, including copyrighted material. This practice has led to numerous lawsuits, with artists, authors, and rights holders accusing AI firms of infringement. For instance, Anthropic, a leading AI company, recently agreed to a $1.5 billion settlement in a class-action lawsuit filed by authors over the use of pirated copies of their works for training.
Also Read:
- Salesforce Faces Copyright Infringement Lawsuit Over AI Model Training Data
- Generative AI Copyright Dispute: Artists and Writers File Landmark Lawsuit Against Google
While companies like OpenAI claim their training on publicly available data is consistent with the US legal doctrine of fair use, and Google-owned YouTube asserts its terms allow for the use of creators’ work for AI model improvement, the issue remains contentious. The US Copyright Office has clarified that works created solely by AI without meaningful human input are not eligible for copyright protection, though human contributions to AI-generated material may be protected. Vermillio’s technology offers a potential pathway to better transparency and compensation in this complex and rapidly evolving digital environment.
