MANAGING THE SORCERER’S APPRENTICE: How AI is disrupting copyright rules around the world
Dr. Angus Finney, experienced producer and international entertainment specialist, shares his insights on the crucial topic of AI and Copyright.
“Whenever I hear the word copyright,” opined one of India’s leading film producers, “I think of it as my right to copy.” This quote always wins a laugh no matter where you are in the world. However, thanks to the rising ubiquity of artificial intelligence (AI), the murky area of what constitutes original work, fair use and copyright infringement etc., is now a deeply serious debate across the global creative industries.
U.S. Law and AI
Earlier this year the U.S. Copyright Office issued a “Notice” stating its position on how and if a work with elements created by AI can be protected by copyright. The Notice went on, buried in the footnotes, to state that only works created by humans qualify for protection, so sole AI elements by themselves don’t qualify. But according to LA entertainment lawyer Schuyler Moore, if we (aka ‘humans’) modify or assemble the AI elements in a way that meets the minimum level of creativity needed for copyright protection, the human co-piloted work can qualify for protection.
So far, that’s highly subjective and hard to pin down. But we can assume that a film that includes AI elements would qualify potentially for copyright protection, even if the AI elements alone would not. The Notice went on to state that copyright registrations needed to identify all AI elements in the work, so that the Copyright Office could determine the scope of copyright protection. Moore points out that some humans may conveniently “forget” to identify AI elements, since it's a tough call for the Office to independently identify them (although down the line AI might just help them out…). The “I forgot” defence won’t hold, but the onus will be on proving the extent of the AI elements in question.
Debates between Creators and Tech Companies
Meanwhile, the WGA and SAG-AFTRA have raised significant objections and concerns based on potential copyright abuse by AI platforms and their corporate owners. The WGA has boldly stated that “plagiarism is a feature of the AI process,” while SAG-AFTRA argues that all rights involving the digital stimulation of a performer and the creation of new performances and recorded performances “must be bargained with the union [and that] “permission or compensation” must be sought across all AI developers’ “use and exploitation of creative works or professional voices and likenesses.”
In response, AMPTP, the management’s negotiating vehicle, has offered a yearly conference-style “chat” about the advancements of technology across the screen entertainment industry. Hence the two sides remain worlds’ apart.
Taking a step back from the picket line, the issues around copyright and AI need to be carefully framed. On one side, AI platforms and their respective owners maintain that when their machines (aka ‘Bots’) scrape the worldwide net for data, images and material, they are simply ‘training’ their machines. The ‘exception’ arguments run the gamut from: i) content is for research and educational use, ii) the information gathered is not directly disseminated to the public, and (the most vague and frustrating), iii) that what AI curates is not infringing any specific copyrights, and that anyway it's impossible to reverse engineer the AI LLM’s (large language models) to discover what it specifically learned and used in the process. That’s not how AI works, we are told.
Legal experts are somewhat sympathetic on the last point: “It’s nearly impossible to pinpoint what is being used given the complexity of algorithms and the billions of data points available on the internet,” explains Steve Saltzman, entertainment lawyer and partner with Fieldfisher. “The argument goes that scraping and training is therefore not an infringement – very much along the lines of the commercial exchange arguments posited by search engines and their owners such as Google in previous decades.”
Rights holders feel rather differently, explains Saltzman. “The copying of text, with no citation, compensation and/or licence, constitutes copyright infringement. And as far as writers go, for example, there are no exceptions. And there are now lawsuits to this effect. However, the results are not in, and there are no precedents here yet, and there likely won’t be for a long time given that most cases will get settled.”
“I think like many things, copyright law did not foresee this day. The amounts at stake are less than the principles. What is going to be critical is some kind of intergovernmental collaboration to agree on standard definitions and uses.”
International Regulations around AI: the UK, EU and China
Not all machines are created equal: Unlike in the US, a machine in the UK can in principle ‘claim’ copyright over its outputs while the current regulation is focused on how AI systems are used rather than policing the technology itself. Likewise, EU law allows for machine generated works to claim ownership of copyright but the detail in practical terms will be enshrined in the upcoming EU AI Act, expected to be fully approved by the end of the year.
“The concern is that there are wide ranges of different objectives, agendas and models currently being used around copyright and IP by different territories. In a perfect world international cooperation that works together to build a framework for a global regulatory system would be preferable,” says Shireen Peermohamed, a partner at Harbottle & Lewis. No one is holding their breath.
China is currently preparing a broad national AI law and has tightly targeted regulation around all new technologies, including generative AI – which must adhere to “core values of socialism.” Recent regulation announced this August catches all products that can “impact public opinion.” Baidu and ByteDance launched new generative products earlier this month with government approval. All the restrictions apply to foreign companies, which make content-generating AI products very hard to launch ‘off shore’ to the Chinese public, while the idea of China signing up to a global regulatory agreement are generally agreed to be a non-starter.
Chinese attitudes to IP and copyright were starting to become more aligned with western values pre-COVID, but have taken significant steps backwards due to increased government interference around censorship. According to producer Zixi Zhang, currently co-producing China’s largest drama series, “Chinese content makers are completely ‘wowed’ by AI and its potential uses creatively, with little consideration if at all about copyright issues,” she explains. “Copying is part of the Chinese business culture, which is obsessed with making money and sees AI as a limitless opportunity.”
A Balanced Perspective on the Unknown
Taking a wider perspective, Dr Chris Mattmann, NASA computer scientist and AI author points out: “It’s key to separate fear and fiction. There are genuine ethical concerns around AI usage and right now different territories have differing [or no] guard rails.”
“We all need to understand that modern AI development has been biologically inspired. The study and application of deep neural networks form the basis of AI progress. The catch is that the top ten thousand AI scientists don’t know exactly how to explain how the machine(s) came up with their outputs. There is not enough learning going on from those AI scientists working on the tech mountain tops.”
Who would have thought that William Goldman’s dictum that “nobody knows anything” was going to be extended to the new machines and the way they work. It leads us to worry about that other famous dictum, the “unknown unknowns” that lie out there, humming away and never asleep on the job.
Dr Angus Finney is an executive producer and Fellow at Judge Business School, Cambridge University