A storm is underway in the usually sleepy world of copyright, and the next bout of lightning is expected in just over a month’s time.
From August 2nd, new transparency requirements of the EU’s AI Act will come into effect. Among other things, providers of general purpose AI models will need to publish a summary of their training datasets, and implement policies to ensure compliance with EU copyright laws around training data, as the law firm Osborne Clarke explains.
With this next step, and as with most things regulatory in nature, you’ll be unsurprised to discover that the US and the EU are going in totally opposite directions on the issue of data transparency for AI - and in turn copyright.
For those who favour regulatory oversight and guardrails, the EU is on the case. The AI Act is steadily being implemented, with various phases rolling out over a phased period. This adds to a recent report on copyright from the EU’s Intellectual Property Office (EUIPO), which highlights that no solution for copyright holders to protect their rights has yet emerged, but is needed in relation to content used for both AI training inputs and AI-generated outputs. Collectively, these are key steps towards establishing a legal framework for how copyright in the age of AI might work.
For those who favour untrammelled innovation around a new type of technology to allow it to flourish, the US is doubling down on that. Just as the US Copyright Office was releasing a new study on AI and copyright last month, President Trump fired both the head of the Office, and the Librarian of Congress (which oversees the US Copyright Office).
The move seems to be related to the US Copyright Office’s “pre-publication” study, which argues that AI companies training their models on copyrighted materials may not necessarily be protected by the legal doctrine of fair use. As NPR summarises, “in some instances, using copyrighted material to train AI models could count as fair use. In other cases, it wouldn't.” Hardly the stuff of fireworks, but still seemingly going too far, given the dismissals.
For those (like me) needing a recap, fair use of a work is defined as being: “for purposes such as commentary, criticism, news reporting, and scholarly reports.” In those instances, you have the right to make use of a copyrighted work without permission from the owner. This gets to the heart of what many tech companies are using as their defence against a host of US copyright infringement cases.
In fact, there are now sufficient copyright-related cases underway against AI companies that US law firm BakerHostetler helpfully runs a case tracker page detailing their progress. One of these cases, on behalf of authors, just this week lost out to Meta, although the ruling left the door open to a clearer and stronger case being made in future. As the FT explains: “This ruling does not stand for the proposition that Meta’s use of copyrighted materials to train its language models is lawful … It stands only for the proposition that these plaintiffs made the wrong arguments and failed to develop a record in support of the right one.”
You can copy me for free
This article isn’t seeking to make the case for or against either the EU or US model. Rather, it’s to highlight an interesting contrast between artists and other copyright holders, and the rest of the B2B world. The one side is (rightly) pushing to have their intellectual property rights upheld. The other side is, essentially, holding up a large sign saying “please plagiarise me”.
For artists whose work is being used without either credit or compensation, this is an intolerable situation and hence the wide range of lawsuits now underway.
But for B2B marketers, who are under increasing pressure to ensure that their companies’ views and insights are the source for any related answers being asked in the millions of daily GenAI queries, being referenced by these AI platforms is a key goal. To use the marketing jargon, they want to ensure they get a sufficient ‘share of model’.
Go to GEO
To achieve this, there’s a major push underway around so-called generative engine optimisation, or GEO. Much like the earlier days of search engine optimisation, B2B marketers are working to update their playbooks on how to get GEO right. In May, Edelman ran an event detailing their approach to how they’re helping organisations get their material referenced within major AI engines. The global PR firm sees it as part of a shift away from driving organic traffic via search engines and towards influencing brand visibility and sentiment on major large language models.
This change in emphasis is not only happening rapidly but might also be a generational shift. In a recent study, McKinsey reports that millennials are the most active users of GenAI, with more than three times as many millennials reporting high levels of expertise relative to baby boomers. My daughter already finds it strange that I suggest Googling for an answer only to get a list of links, most of which have paid to be there, rather than using a GenAI tool that neatly sums up the answer she’s looking for. Welcome to the ‘zero click’ era of search results.
Edelman’s recommendation on how to get GEO right is summarised as: focus on structured multi-media content, earned media coverage, and authoritative organic content. Some of this is likely biased towards Edelman’s strengths (their CEO believes the AI-led search era will lead to ‘the golden age of earned [media]’), but the points are valid. Importantly, these attributes also play to the strengths of B2B marketers that invest in high-quality thought leadership: creating authoritative organic content that also delivers strong press and media coverage. In turn, this content will also boost your visibility within AI platforms.
So what does authoritative content look like? My longstanding business partner and former FT Longitude co-founder, Rob Mitchell, tackled this in a recent article, reporting that some excellent recent academic research shows that “original, well-researched, evidence-backed and well-written content will perform well on AI search tools”. These are all characteristics of the kinds of good thought leadership content that we’ve encouraged companies to adopt, and so they should in turn enable those brands to perform well in this new era.
Meanwhile, for the authors, artists, musicians and other copyright holders out there, the frustration over copyright infringement continues. While the market awaits new legal precedents that could help them to defend their position, they’ll no doubt be looking at distinctly different advice: how to block GenAI from scraping your content.
Photo by Ria on Unsplash