AI copyright laws are on the way to protect machine authorship

To grant authorship to AI, it is highly important to understand how the AI systems are trained

The deliberations around whether AI is given intellectual property rights have rather become chasing a mirage. Recently US Court of Appeals of the Federal Court, in the popular Thaler Vs Perlmutter case, held that artificial intelligence (AI) cannot be named as the inventor of a patent. The US court categorically mentioned that the term "individual", in the US Patent Act refers to human beings.  Now, this comes as a discouraging move toward overall AI development. AI enthusiasts, particularly those who particularly swear by their unique creations or rather the unique abilities of artificial intelligence, are mostly of the opinion that not giving AI, authorship rights, would only hinder R&D efforts. While it is the take of US courts, WTO's 1994 TRIPS agreement specifies that an invention must meet certain requirements of which it being a novel invention is one, and experts should consider it to be a non-obvious outcome. Apparently, law authorities across the globe are left to find answers to some difficult questions, viz., who owns the IP, if AI can change the nature of the invention, etc. Copyright experts are very clear in their opinion that AI is considered a category in patent law for granting IP rights. So far, only countries like the UK and Ireland permit copyright protection, granting authorship to the human who facilitated for AI to make innovation.

What makes AI's work IP worthy?

Being trained on deep learning and reinforcement learning architectures, it learns and adapts to be able to generate compositions. Therefore, it is the training process that should be put under scrutiny when it comes to deciding on whether the innovations are novel. For any functional or creative innovation, data that is used as input is way more important than what lies in the black box, which can be put under an entirely separate issue for discussion.  The data used to train AI systems might give rise to rights issues, licensing issues, and other copyright considerations, given the fact that data is user generated and most of the time it is the end product of the toil and sweat they put into their work. It might seem unreasonable to grant copyright to AI creation that is based on others' intellectual inputs. But a few pieces of international legislation provide exceptions. Japan grants reproduction rights to the extent that they can be used for TDM (Text and Data Mining) without authorization as long as the rights of the right holders are not violated. In the EU, in the case of TDM, the law applies until the rightsholder has not reserved their rights about TDM. Individual laws aside, the majority view leans towards giving copyright authorship to an artwork, provided individual authors have made a unique contribution. For example, the Cosmo magazine cover depicting a female astronaut had an entire team experimenting with different prompts before the final one was chosen. Imagine a case where AI was human, then who would be eligible for the copyrights? One may argue AI cannot have a moral right over what it is trained for, but laws are gradually evolving to take AI's agency into account though with some caution.