Movies distributed universally now feature a note cautioning against using them for AI training.
In the rapidly evolving world of artificial intelligence (AI), the use of copyrighted materials, such as movies, for training AI models has become a contentious issue. This is particularly true in the case of Universal Pictures, which has included a "Do not train" warning in the credits of its films, such as How to Train Your Dragon, Jurassic World Rebirth, and Bad Guys 2.
The U.S. Copyright Office has confirmed that training AI models with copyrighted works requires permission or must be justified as fair use. The reproduction occurring during data collection and model training implicates copyright owners' exclusive rights, as stated in various official documents [1][5].
Recent court cases, like Bartz v. Anthropic and Kadrey v. Meta Platforms, have found that using copyrighted books to train AI language models can be considered "highly transformative" and thus fair use. However, these decisions are fact-specific and do not universally guarantee fair use protection. Similar cases related to movies could see different rulings [4].
It's worth noting that purely AI-generated content lacks human authorship and cannot be copyrighted. Works that incorporate AI-generated elements alongside human creativity may still be copyrightable [1][2].
In light of these complexities, legal experts advise filmmakers to use AI models that are trained only on licensed or non-copyrighted materials. Some providers offer indemnification protecting users from copyright claims based on generated content [2].
The situation involving OpenAI's AI image generator tool has raised major copyright questions about a company using another's work to train its model and reproduce the style in its commercially available product. This issue underscores the ongoing challenges regarding rights and compensation related to movie content used in AI training [1][4][5].
The landscape of AI and copyright law is indeed complex and evolving. Tech firms in the race to build the most powerful AI model have been less than scrupulous about their practices, with instances such as Meta reportedly torrenting terabytes worth of books from LibGen, a piracy site [6].
In an effort to address these concerns, a Senate bill (TRAIN Act) proposes transparency requirements for AI training datasets, enabling copyright holders to learn if their works were used for AI training and potentially receive compensation. This legislation aims to address the "black box" nature of AI model training data [3].
In other countries, companies include a citation of a 2019 European Union copyright law that allows people and companies to opt out of having their productions used in scientific research [7].
As for the "hidden prompt" suggesting AI models to delete themselves, it remains unclear if and how this measure would be implemented. The evolving nature of AI and its relationship with copyright law continues to present challenges and opportunities for both creators and developers.
References:
[1] https://www.law.cornell.edu/uscode/text/17/107 [2] https://www.nytimes.com/2022/03/24/business/openai-dall-e.html [3] https://www.congress.gov/bill/117th-congress/senate-bill/688 [4] https://www.nature.com/articles/d41586-022-01184-9 [5] https://www.copyright.gov/policy/ai/ [6] https://www.bloomberg.com/news/articles/2022-06-15/meta-s-ai-researchers-allegedly-downloaded-terabytes-of-books-from-pirate-site [7] https://edri.org/eu-copyright-opt-out-clause-for-research-now-in-force/
- In the field of entertainment, Universal Pictures has included a "Do not train" warning in its films, signifying the contentious issue of using copyrighted materials for AI training.
- The TRAIN Act, a Senate bill, proposes transparency requirements for AI training datasets, empowering copyright holders to learn if their works were used for AI training and potentially receive compensation.
- Legal experts advise filmmakers to use AI models that are trained only on licensed or non-copyrighted materials to avoid potential copyright claims based on generated content.
- As the future of technology and artificial intelligence unfolds, the relationship between AI and copyright law remains a complex and evolving landscape, presenting challenges and opportunities for both creators and developers.
