A paper written by the European Parliament concludes that using creative material for AI training does not count as fair use, and that “timely reform” is required. For the past couple of years, as AI has begun to dominate headlines, tech firms such as OpenAI have maintained that the use of creative work to ‘train’ ... No, generative AI does not count as fair use, EU report finds
A paper written by the European Parliament concludes that using creative material for AI training does not count as fair use, and that “timely reform” is required.
For the past couple of years, as AI has begun to dominate headlines, tech firms such as OpenAI have maintained that the use of creative work to ‘train’ their software doesn’t infringe existing copyright laws. Platforms like ChatGPT scrape authors’ work from the internet and use it as the raw material for their outputs. So when you ask the chatbot for, say, a recipe for chocolate brownies, it isn’t a sentient computer coming up with the correct ratio of ingredients, but rather a piece of software synthesising a set of words based on other people’s online cookbooks.
This is, according to tech firms, sufficiently transformative to count as ‘fair use’. In a potentially significant moment in the ongoing debate, a report commissioned by the European Parliament has concluded that those firms are wrong.
The report, entitled Generative AI and Copyright, was published only a few days ago, and brought to our attention by The Register. It’s a lengthy manuscript at 175 pages, but its executive summary provides a glimpse of its findings.
“While innovation is nothing new to copyright law, generative AI presents an unprecedented test of scale, opacity, and economic impact,” one section reads, before adding that, “without timely reform, the EU risks legal uncertainty, market concentration, and cultural homogenisation.”
The report then refutes the ongoing argument from tech companies that their AI products are simply learning as humans do – that they assimilate information and then create new work based on it. Here’s the quote in full:
“While it is often suggested that AI systems ‘learn’ in ways similar to humans – such as reading a book or studying a painting – this analogy is misleading from a legal perspective. Under EU copyright law, this study finds that such a comparison does not hold. When generative AI models are trained on protected content, they typically make copies and process the actual expressions found in those works. This goes beyond what is permitted under current legal exceptions for activities like research or analysis.”
Read more: With its AI action plan, the UK government risks selling us out to chatbots
The report the recommends that existing laws need to be bolstered in order to protect creative and media sectors from being ransacked by AI companies, and that those whose work has been used should be paid. Or, to put it in cooler terms:
(2) Fully machine-generated outputs should remain unprotected; AI-assisted works require harmonised protection criteria.
(3) A statutory remuneration scheme is essential to bridge the growing value gap between creators and AI developers.
The findings echo what many people on the side of those creative and media sectors have been saying for months. But at the same time, it’s looked as though some governments have listened to AI firms instead; in the UK, the government has decided that tech firms should be allowed to use copyrighted material for training purposes. In the US, OpenAI is leaning on the Trump administration to conclude something similar.
Globally, the legal system is rushing to catch up with AI’s implications. Meta recently won a copyright case over the use of authors’ books for its AI models. Meanwhile, Disney and Universal is suing Midjourney for copyright infringement, calling its image generation software a “bottomless pit of plagiarism.”
As the EU report points out, though, generative AI is already having a massive impact on publishing of all kinds. Thanks to Google’s AI results and ChatGPT, traffic to websites is tumbling, affecting vital advertising revenues. This latest report might be a sign that, amid all the fear and loathing, there’s hope that creators and artists of all kinds will eventually find some kind of protection from the storm.
You can read the full report in its lengthy report (PDF) on this EU website. Maybe make a cup of tea before reading.
