Will the EU’s AI regulation slow agencies on the continent?

The letter argued that the Act would “jeopardize Europe’s competitiveness and technological sovereignty.”

Some of the most advanced efforts to experiment with AI tools, or integrate them into the everyday work of advertising professionals, have come from agency businesses based within the EU. Ogilvy’s AI Lab, for example, is located in Paris. Media.Monks and Dept – both bullish about AI’s business potential – were founded in the Netherlands; the former's business now sprawls across German, France, Italy, Spain and Sweden.

Dept’s digital ethics ambassador Lukas Stuber says he’s optimistic about the regulation in its current form. “I am really, really happy about the direction that the European Union is taking here,” he says. “I think they are raising exactly the right issues that we as a society, and we as an agency with our clients, will need to address in the future.”

The EU’s proposals will take a long time to become law across the continent – a final draft may be ready by the end of the year, and formal ratification could take until 2027 – and they are extensive.

The draft Act proposes banning outright real-time and remote biometric ID systems, ‘social scoring’ systems and AI tools that enable the “cognitive behavioral manipulation of people.” Content created by generative AI tools would have to be labeled, and the large language models (LLMs) underlying those tools would have to be designed not to generate illegal content. It also provides a brief risk framework, meaning that AI systems being used in, for example, law enforcement or the educational sector would have to be registered in an EU-wide database.

Risk framework

The impact of that risk framework will primarily hit agencies with client rosters that include more of those ‘high-risk’ industry players. The EU’s proposals would deem financial services, education, healthcare and public services ‘high risk.’

Dept’s vice-president of emerging tech Isabel Perry says more than a few of its customers would be affected. “The way the AI Act is set out, it will affect some industries more than others. We estimate a third of our clients would be immediately impacted by it,” she explains.

Staying in line with those rules might mean avoiding specific AI applications or pausing work already underway. However, Stuber and Perry say that Dept’s services would already fall within the guidelines set out by the EU.

The impact is less clear for agencies that aren’t as heavily leveraged toward the financial services sector. Wesley ter Haar, Media.Monks co-founder and now executive director of parent company S4, tells The Drum: “The risk framework seems focused on more high-risk systems and applications, which don’t necessarily translate to marketing, but we’ll be following it closely.”

He’s more concerned with the regulation of training data. The Act limits what data can be used to train the enormous datasets behind large language models (LLMs). The topic has already generated several lawsuits, including a huge one against Google’s data-scraping practices.

Ter Haar says he’s “very interested in how the general legality of training data discussion plays out.”

He adds: “There are quite a few lawsuits being fired up at the moment, so we should see more clarity there hopefully soon-ish.”

There’s also the question of liability. Who’d bear the blame if a specific tool or application breached regulations? And will that consideration slow down agencies that want to invest in their own generative AI solutions?

According to Stuber: “Liability will quite probably be placed with the tool owners, as opposed to bespoke solutions, where liability will be firmly placed on our shoulders.”

Gold standard

The EU’s AI Act is the first major piece of legislation on the topic, and if it isn’t significantly diluted, it will also be one of the farthest-reaching. Because of that, Stuber expects that Dept would adopt the EU’s rules as standard in every territory it operates in. “We see a lot of parallels to GDPR,” he adds.

Ter Haar says Media.Monks would likely follow a similar approach. “With previous EU regulation, like GDPR, it became the standard adherence for us in our global practice, the same could happen with AI, but in any case, we’ll be following rules and regulations closely in every region we operate,” he says.

Though it might take a long time for any of those measures to come into effect, Stuber says that Dept is attempting to get ahead of regulation by focusing on transparency, copyright and ensuring that AI tools don’t contribute to bias. “In my opinion, we have to adopt the direction of the European Union quite quickly. It’s not enough to publish some fluffy code of conduct,” he says.

“By adding transparency, self-imposed or via watermarking... you deliver a very important data point to the audience. It’s important to gauge the veracity of the content they are consuming. We will be absolutely obliged to do that.”