The Verve’s track “Bitter Sweet Symphony” was a huge hit around the world. Nominated for and/or winning various accolades, it is considered one of the defining songs of the Britpop era. And it was caught up in what many consider to be one of the harshest copyright conflicts in modern music history. To cut a long story short, the dispute boiled down to how much of a sample of an orchestral version of the Rolling Stones song “The Last Time” (written by Mick Jagger and Keith Richards) was used by the Verve in “Bitter Sweet Symphony.” Despite the Verve having obtained what they thought was an appropriate licence, until 2019 the 1997 hit was credited to Jagger and Richards rather than the Verve’s frontman Richard Ashcroft, denying Ashcroft millions in publishing revenues.
All of this happened in plain sight. The Verve didn’t hide the fact they were using a sample of an orchestral version of “The Last Time” — they even tried to obtain the correct permission — but still the law sided with the holders of the copyright of the original content.
The law is having a tougher time protecting the rights of content creators from copyright infringement by artificial intelligence (AI).
Let’s say you’ve written a short story with an ingenious premise. And someone takes the premise and the bones of that story and turns it into a novel, which becomes a bestseller, and the rights are sold for six seasons and a movie, and the deal for the rights includes a hefty cut of all merchandise. But you see none of the novel/series/movie/merch money. You’d be cheesed off. And rightly so.
Now imagine that happened to everyone who ever created a thing. That’s where we are with today’s AI industry.
Data Grab
In the early days of large language models (LLMs), tech companies trained them primarily on publicly available data sourced from the internet. This included books, Wikipedia articles, online forums, academic papers, and open-access news sites. The rationale was that vast amounts of text were needed to develop models capable of understanding and generating human-like language. At the time, there were few legal or ethical restrictions on using publicly accessible content — web scraping and dataset compilation became common practices in AI research.
But as models grew larger and more sophisticated, concerns arose about copyright infringement, data privacy, and content ownership. Many creators and publishers objected to having their work used without consent, arguing that AI companies were profiting from their intellectual property without compensation.
The big problem for creators is that it is extremely difficult to prove that AI companies are pirating content, as they delete their data sources after feeding it to their LLMs.
AI’s Cost
The Industrial Revolution changed the fundamental relationship between labour and capital. We are soon to see a repeat with the AI Revolution, but the impact will be psychological as well as economic and political.
Global management consulting firm McKinsey & Company predicts that 30% of US work hours will be automated by 2030, which will displace 12 million workers. That’s just the USA, in just five years. What will the impact be worldwide?
And then there’s the issue of compensating data creators for their contributions to AI training.
What Can Be Done?
Various initiatives and proposals aimed at ensuring fair recognition and remuneration for data creators have been discussed. One notable development is the formation of the Dataset Providers Alliance (DPA) — a coalition of AI data licensing companies — which advocates for standardised and ethical AI data licensing practices, emphasising an opt-in system that requires explicit consent from creators before their data is used for AI training. This approach contrasts with current opt-out methods employed by some major AI companies and aims to establish fair compensation structures, such as subscription-based, usage-based, and outcome-based licensing models.
And the International Monetary Fund (IMF) has also weighed in on this issue, advising governments to consider a number of measures to mitigate the impact of AI, including a carbon tax to take into account the environmental impact of operating the servers that train and operate AI systems.
But it’s important to remember that AI will bring benefits; our next moves shouldn’t be about stopping AI development. We need to ensure its benefits are distributed as widely as its costs. AI’s unparalleled acquisition of content needs to be transformed into the greatest redistribution of technological benefits in history.
