What legal challenges does the fashion industry face in the era of generative AI?

From safeguarding intellectual property to ensuring its own use of artificial intelligence, the fashion industry is still finding its feet with AI. As expected, the topic took center stage at the Assises Juridiques de la Mode, du Luxe et du Design, held in Paris on December 9 and organized by Lexposia.

“In 2024, we will submit 2.5 million reports of counterfeit content to platforms,” ​​explained Nicolas Lambert, director of online brand protection at LVMH. “This is nothing new, but AI has made it increasingly easier to generate infringing content. Currently, for example, we are seeing a proliferation of online ads for counterfeit Advent calendars from Sephora, Dior and other group brands.”

Alexandre Menais, general counsel of the L'Oréal group, was also present to testify to this acceleration. In his opinion, the increasing presence of this new technology requires reflection on the interactions between the company and the machine and, in particular, on how these interactions are used.

“With an intelligent agent, the question arises of who owns that interaction,” the legal expert stressed. “One of the risks I see is that the rules set by companies, which require the use of closed AI, will be widely flouted. Many employees will be tempted to try AI outside the established framework.”

Christiane Féral-Schuhl, a lawyer specialized in this area, also identified this risk. For the former president of a bar and former president of the National Bar Council, it is urgent to raise awareness among employees about the differences between a closed AI, formed on creations and data on which rights holders have given their consent, and an open AI system. The latter dispenses with the consent of rights holders by relying on the “text and data mining” (TDM) exception.

“These AIs are ogres that swallow all this 'training data' and to counter this, you can build your own AI system, using protected data within a controlled framework. If an employee prefers to use an open system, they feed the machine and in effect share their work and creations with others (including their competitors) who can exploit it to produce infringing works.”

Féral-Schuhl also highlighted the questions that need to be asked of AI tool providers. Some stipulate in their terms that one client's work can be used to improve service for all clients, which, in a creative context, should obviously be prohibited.

Frédéric Rose runs IMKI, which designs custom generative AI for brands like The Kooples and G-Star. The specialist points out that AI is increasingly sophisticated. “You will soon be able to write patterns and technical execution files,” he estimates. “It is already becoming more and more precise and able to specify materials, fabric weights (grammage) or sewing types.”

This level of detail now makes it possible to detect counterfeits, both for rights holders and consumers.

“Some AIs have safeguards and refuse to respond, but others give suggestions on where to find the best dupes,” Lambert said. “Between the AI ​​and the customer, it's a private channel that I can't investigate. But maybe tomorrow the AI ​​can identify suspicious behavior. Maybe we need to imagine, as with YouTube, a DMCA-style mechanism (a rights holder takedown mechanism, editor's note) that prevents an AI from directing users to a counterfeit product.”

“And if AI is exploited for creative purposes, it is also necessary to define red lists of iconic elements, specific signatures, that could make a creation resemble that of an established brand,” said Féral-Schuhl.

It also points to the appearance of “watermarks” (or digital tattoos) of the data used to train AI, which could eventually be subject to copyright protection and prevent their use in the creative processes of AI agents. This is in addition to “information tagging” that records the date and location of AI-generated creations.

For his part, the vice president of the French unicorn Mirakl, which develops markets for large retailers, Hugo Weber, spoke about the contribution that AI could make to algorithms that are already highly efficient.

“Amazon Prime is not a logistics issue: if they deliver to you the next day it is because in 95% of cases your purchase was already being shipped, because the algorithm is very efficient,” summarized the specialist.

He also warned against turning the Shein case into a test of markets, noting that European, American and Chinese actors have different notions of responsibility.

The Shein case was also addressed by Benoît Loutrel, president of the working group on online platforms of ARCOM (Autorité de Régulation de la Communication Audiovisuelle et Numérique).

“We are moving from preventive action by regulators to coercive action by the courts. I think the next stage will be civil law, especially in the case of artificial intelligence,” said the specialist.

With ARCOM equivalents rising in other European countries, he hopes to see French digital sovereignty anchored in the broader European Union framework that is now taking shape.