Artificial intelligence is hotly debated in the realm of copyright. The rise of generative AI, producing text, art, and even music, has ignited a legal firestorm. Is AI a threat, ushering in a copyright apocalypse? Or is it just another tool to be capitalized on? The answer (as with most things in technology) is far from simple.
Generative AI: Copyright Disaster In the Making
For many creators and rights holders, the very existence of generative AI feels like a robbery.
In case you have been living under a rock, AI models are trained by ingesting huge datasets, usually scraped from the internet, that always include large amounts of copyrighted material. Media companies argue that this constitutes one massive, unauthorized, uncontrollable copyright violation.
The fear is that AI, trained on the scraped data, can generate new content that competes with the original. This would destroy the traditional market and generally devalue creative work.
Imagine an AI generating the next Song of Ice and Fire book. Imagine a brand new album by Freddy Mercury. Economically and culturally, the consequences could be devastating.
And the fun part is, this scenario is no longer theoretical. Major players are taking legal action.
Disney and Universal, for example, recently filed a lawsuit against AI firm Midjourney, alleging extensive copyright infringement. Their 110-page complaint details how Midjourney’s image generator allegedly “stole countless copyrighted works” to train its system, resulting in outputs that mimic characters like Darth Vader, Elsa, and Shrek.
Similar lawsuits have been brought by the New York Times against OpenAI and Microsoft, and by Sony Music Entertainment against AI song generators like Suno and Udio. The latter is especially interesting, as the outcome is about, according to Forbes, “setting the rules, or perhaps abandoning them, for how copyrighted music is used in training AI”.
What If It’s Not Such a Big Deal?
On the other hand, some argue that the “copyright apocalypse” narrative might be overblown by media companies.
For one, building upon existing work is fundamental to art. Writers, painters, and film makers are naturally influenced by those who came before them.
Echoing this sentiment, the legal doctrine of fair use is often brought up as a defense for AI. Many argue that training an AI falls under “fair use” as transformative because it’s not simply reproducing the original. It’s extracting patterns, styles, and concepts to create something new. The output, they say, is a new expression, not a copy.
(Meanwhile, the US Copyright Office has stated that while purely AI-generated works are not copyrightable, works that involve “sufficient human authorship” in the selection or modification of AI-generated material may be protected. So there is that.)
The battle goes on.
Nothing New Under the Sun
If you think about it, technology has presented copyright challenges before, and the legal system adapted. The internet, social media, and digital distribution platforms have all, at various points, been accused of copyright infringement on a massive scale. No-one died. The corporations are okay.
In fact, Facebook (Meta) itself has been caught scraping pirated book archives to train their AI model.
Now that the cat is out of the bag, you cannot make it climb back in.
One thing is clear: AI tools can democratize content creation, just like smartphones did on a smaller scale. Too many regulations could stifle the potential of AI to create new forms of art. Lax regulations could harm professional creators. The truth is in the middle, and the next couple of years will establish enough precedents.
More like this:
- Are You Safe From the AI Web Scraping?
- Running AI Locally: The Pros, Cons, and Popular Methods
- Hunter S. Thompson: Daily Routine & Writing Process
