Last week, over 8,000 authors who have published works sent an open letter to the founders of a generative AI platform. They sent this letter to the CEOs of major tech companies, calling for fair compensation because these companies use their copyrighted works to train their models.
This conflict is another intellectual property issue that plagues AI development. In another interesting case, a US district judge in Northern California dismissed most of the lawsuits filed by a group of artists. He demanded more evidence regarding the alleged copyright infringement from Stability AI's related source code.
The issue with copyright is not unique to generative AI. Throughout history, copyright laws have undergone major revisions with the emergence of any new technology.
Pamela Samuelson, the Chief Scholar of Law and Information at the University of California, Berkeley, raised three key questions regarding generative AI. First, does using works as training data for generative AI systems constitute copyright infringement? Second, when does the output generated by AI constitute derivative works, and who owns the copyright to the computer programs and outputs, which are protected by copyright? Lastly, who currently owns this copyright?
These questions address the fundamental legal and ethical complexities surrounding generative AI, which impact the rights of original authors and the expanding field of AI-based innovation.
In January 2006, a court ruled that the act of search engines copying copyrighted works on the internet for indexing purposes falls under fair use rather than infringement. The Field v. Google case was significant because the court considered digitizing millions of copyrighted books in libraries and providing excerpts in response to search queries and other computational uses as fair use.
Google did not utilize the expressive aspects of these works but rather provided a summary when asked. Stability AI and other defendants use these cases to argue for "fair use" since web scraping content from the internet for training purposes is considered fair use rather than infringement.
"We believe these large tech companies are preparing for larger datasets they can use," said John Degen, Executive Director of a Canadian organization. "We are right." There is a distinction between making it convenient for people to find the owners of copyrighted works and ingesting copyrighted materials for training data. Artificially created images or texts compete with the original data, and the authors of the content did not agree to this.
According to the open letter, they believe it is fair to receive fair compensation because the value of the ingested materials is what enables generative AI to produce outstanding results. From this perspective, the argument is highly persuasive. Otherwise, generative AI systems would produce garbage. Based on this viewpoint, carefully crafted works should be paid for by someone.
If copyright law only concerns itself with protecting the original expression in works, that should be a factor because those who use works as training data do not consider them from an expressive standpoint. They consider them more from a computational use and text data mining perspective, which has historically been considered fair use.
Generative AI systems can create new works, and the purpose of copyright law is to promote progress in science (creation) as the founders refer to it as knowledge and practical technology. It can be argued that generative AI systems advance this purpose, and fair use provides some space for new creations, which may also involve whether ingestion is considered fair use.
In April of this year, Japan confirmed that their existing laws allow for the use of data collected from the internet for non-commercial and commercial purposes. Japan did not introduce new policies, but since 2018, copyright holders have been able to utilize existing exceptions for text and data mining. Content used from illegal websites is subject to claims, injunctions, and criminal penalties, but it is difficult to prove that language models actually scraped these websites without the company's admission. Israel follows similar principles but with some exceptions. Specific datasets cannot be used to train AI models to generate works that are very similar to original works, such as using the "Game of Thrones" series to train an AI model to generate very similar works.
Germany also stands on the side of technology and refuses to respond to the concerns of over 140,000 writers and artists. The government believes that there is no need to strengthen regulations. Digital policy spokesperson Maximilian Funke Kaiser said, "Publishers and media companies also benefit from this technology, for example, through AI-supported text generation."
On the other hand, the European Union requires companies deploying generative AI tools to disclose any copyrighted materials used in developing their systems. This is the most basic requirement and is considered fair. However, as mentioned earlier, the process of proving this is difficult and almost impossible. But according to Pamela Samuelson, this aligns with fair use in copyright law. Training AI on copyrighted materials that are protected and not creating derivative works and then generating non-derivative original characters through AI is permissible.
While you cannot copyright a particular anime style, unique art styles and narrative approaches belong entirely to the original works and cannot be completely replicated. So, which companies are being sued for copyright infringement? And why is it so difficult to share details about the training data?
Source: https://analyticsindiamag.com/the-copyright-tussle-ai-companies-are-sure-to-win/
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.