As Andy Warhol famously said: “I don’t know where the artificial starts and the real stops”. In the age of artificial intelligence, such cryptic words resonate prophetic – especially considering that AI itself allowed the pop-art star, who passed away long ago, to narrate his own Netflix 2022 docuseries, "The Andy Warhol Diaries". 

The relationship between artificial intelligence and audiovisual works, however, goes far beyond the reproduction of the voices of famous people from the past. For instance, the BBC’s wildlife documentary series, "The Watches", have been using computer vision software for years to film the most striking moments of British wildlife. More recently, a number of top-tier broadcasters have lately begun airing AI-anchored newscasts and economic programs. But there is more: for some time now leading television production companies have leveraged artificial intelligence to try and generate original format concepts. For example, in August 2023, French multinational TV production company Banijay unveiled its AI Creative Fund, which encourages exploring the endless possibilities of artificial intelligence when applied to content creation. When announcing the initiative, chief content officer James Townley remarkably clarified that the fund will be used to merely “empower producers” and that “human creativity will always prevail”.

But when it comes to the law, the relationship between human and “robotic” creativity could be a little more complicated than that, especially as regards the thorniest legal question arising from the use of AI systems to create original TV works: Who owns the rights in the format? 

As long as the criteria of originality, creativity and expressive completeness are met, TV formats can access copyright protection. But when the concept stems from the use of AI systems, the waters get muddy; at present, no statutory provisions in the EU expressly address the ownership of IP rights in AI-generated works.

Nonetheless, some guidance can be sought in the many non-binding commentaries and opinions dealing with this issue, especially "Trends and developments in artificial intelligence", a report recently published by the EU Commission. It points out that the answer to our question should depend on the role played respectively by the two key (human) players involved in AI-driven content generation: the AI system user and its developer. In fact, since copyright protection arises from human creativity and copyright laws only apply to human beings, the relevant rights (if any, since originality is always required) can vest either in the developer or in the user of the system. Accordingly, the greater the user’s efforts, the more likely he will be deemed the author of the AI generated content – and consequently its owner, since generally ownership follows authorship. 

On the other hand, if the user’s role is relegated to an initial prompt or even merely pushing buttons, the AI system developer would probably qualify as author and owner of the AI-assisted outputs. In cases where the roles of user and developer are instead of equal or similar weights, co-ownership of the results might be envisaged.

Last, but notably, it must be considered that many of these doubts are likely to be prevented by what the EU Commission calls “obvious commercial reasons”. Indeed, let’s imagine that a TV production company appoints a software house to develop an AI system to be used for conceiving original TV formats. Should there be any risk the AI developers would eventually raise claims as to the ownership of the system outputs, the TV production company will probably turn to another provider. Contractual T&Cs for the use of AI systems will arguably address the problem by precluding any authorship and/or ownership claims by developers and ensure that the system users retain all the relevant rights.

But some questions remain unanswered. As AI-enabled autonomy continues to grow and develop, artificial intelligence systems are increasingly able to “train themselves”, gradually achieving results in relation to which human inputs (from both the user and developer) may shrink to the point of becoming negligible. In such a scenario, the above legal landmarks may no longer apply, and the European copyright system may be forced to make a paradigm shift: Will machines become eligible for authorship? And if so, who will own the results of their “artificial creativity”? This bring us back to our original question: Where does the artificial begin, and where does the real stop?

In this dynamic intersection of artificial intelligence and copyright law, we will continue monitoring the evolving legal landscape surrounding AI-generated TV works and analyzing how the law will shape the authorship of AI creations.