[Guest post] Conference Report – Fashion Reborn: The Age of AI

It was with great pleasure that I was able to attend Fashion Law London’s latest event held on 13 May 2024 focusing on the impact which AI innovation is having within the Fashion industry. The well-attended event was chaired by PermaKat Eleonora Rosati and Giulia Gasparin. Attendees were treated to an afternoon of deeply interesting discussion on how artificial intelligence is permeating the fashion industry and what this ultimately means for rightsholders, AI operators and the state of the law. The sessions provided insight from technical, industry and legal experts (not necessarily mutually exclusive) about what the key issues cropping up and what the future holds.

In the course of the afternoon, it became quickly clear that fashionistas are very much part of the technological revolution as the industry races to innovate with the law following in second place. The pace of change provides some uncertainty about what the future looks like for AI within the fashion industry and its potential applications. Beyond the lack of clarity on the practical applications, this uncertainty spills over to how the law treats (and will treat) the use of AI. Prospective legislation is far from perfect with issues arising from unclear language and overlapping regimes. The lack of certainty does not, however, provide businesses with a carte blanche to sit on their hands and wait for the answers to appear from the fog. Business will need to think critically about their practices and simultaneously keep an eye on the introduction of various pieces of legislation in the coming years and evolving case law.

Whilst this is enough to make the average Katfriend want to resort to retail therapy, the panellists provided a practical insight and enjoyable discussion on these issues. The afternoon was broken down into three sessions:

1. Mapping uses of AI in fashion
2. The IP dimension of AI in fashion
3. The regulatory, commercial and data protection dimension of AI in fashion

We first hear from panellists Kirsty McGregor (Vogue Business), Natalie Varma (Farfetch), Paul Jurcys (Prifina) and Femi Idowu (Microsoft), who provided for a multi-faceted discussion from both fashion and tech about what AI in fashion looks like and how it is expected to develop.

Mapping uses of AI in fashion

We heard from the panellists that the revolution was only just beginning as far as AI is concerned. There was agreement that the use of AI is not a new technology to the fashion industry. Rather, the focus on AI has merely intensified as a result of relatively new technologies, such as ChatGPT, appearing on the scene.

This of course is in-keeping with general trend within the fashion industry to use data points to better understand their consumer and drive efficiencies. The panellists shared that this trend was set to continue with the scope of possible data collection points increasing dramatically. This is being made possible in part through new opportunities for device integration into garments or accessories where we heard examples of clothing providing feedback of how the wearer is performing in sportswear or (for the less active among us) even when your favourite jacket needs to be dry-cleaned.

There was agreement amongst the panel that there is likely to be a material difference in consumer’s demands and expectations of how AI should improve their experience. Part is a shift in local processing of their data made possible through AI. The other major shift is likely to arise from consumer’s expectations of how their shopping experience can be personalised both in store through an increased screen footprint and online.

However, it seems that the changes are not just destined to be on consumer-facing applications. Behind the scenes, production methods are ripe for improvement with AI-lead data analytics aiding to improve the efficiency of existing methods. Another example is the generation of website imagery or content which has traditionally been a somewhat time-consuming (and undoubtedly slightly repetitive) manual task. Some of the panellists felt that the focus of the industry has so far been to provide the “next best” version of current industry practice. It was felt that AI’s full potential would only be made possible with a complete review of the industry.

There is also an opportunity for brands to become more sustainable through the adoption of AI. Looking forward, there is scope to drive efficiencies with manufacturing practices and through enhancing a brand’s ability to better predict buying and merchandising needs. This is in turn can only improve the consumer’s experience to ensure they can get what they want, when they want and providing them with better data to ultimately ensure that they are going to want to keep it. Each of these aspects provide a promising future in terms of reducing waste and inefficiencies within the fashion industry.

The question remains whether this is achievable for all brands. Given the level of investment and data resource needed to develop and train AI models, there is a potential barrier to entry for smaller houses or brands that don’t have the base-level or resource. This could pose an issue for brands as consumers’ expectations evolve to expect an AI-enhanced shopping experience.

Next Sir Richard Arnold LJ, Benoit Van Asbroeck (Bird & Bird), Spyridon Sipetas (Fashion Law London), and Xuyang Zhu (Taylor Wessing) delved into the ways in which the legal regimes in both the EU and UK are adapting (or perhaps not) in response to the increased use of AI. This was considered from the perspective of the different phases of use for AI: input, training, output and then any potential liability.

The IP dimension of AI in fashion

The discussion on the input phase began whether the use of certain source information to train AI models could infringe third party rights. A prevalent theme of this discussion was that the answer to this question would likely depend on the training models and whether the process by which it was accumulating and then using the information once scraped. Naturally at this point, the focus would then need to turn to any applicable exceptions under CPDA 1988. Current exceptions under the Act for computational processes are unlikely to be suitable given the preclusion of commercial activities and need for lawful access. An attempt by the Government to modernise the law in this area has been unfruitful. The burden is therefore placed on the courts to triage these issues for the time being with the definitive outcome of Getty Images v Stability AI being eagerly awaited.

Turning to training, a different perspective was offered from the current state of law within the European Union. In particular, whether the introduction of opt-out rights for rightsholders under article 4 of the Digital Single Market Directive can offer any safe harbour for rights holders seeking to protect their works from being subsumed into AI modes.

Considering output, the panel reviewed recent litigation looking at the question of how authorship can affect the subsistence of rights. A particularly interesting point of debate arose on whether the question of authorship might be putting the cart before the horse given that both copyright and unregistered design rights require some degree of originality. This requirement connotes some degree of choice or decision making which overlays an interesting paradigm on machine-created works.

For the time being, there is no indication of how these issues will be resolved in England & Wales. Some indication of how these decisions might be treated are emerging form the US Copyright Office, where decisions from registrations of AI generated works could provide a framework for discussion for European jurisdictions.

Last and by no means least, liability was considered from the perspective of use case. Most brands will likely be deployers of AI technology source from suppliers. As such, they should be mindful of their obligations arising from the EU AI Act.

In the final session, Francine Cunningham and Adriano D’Ottavio of Bird & Bird and Giulia Gasparin hosteddiscussed how the use of artificial intelligence is regulated by new legislation as well as existing data protection regimes.

The regulatory, commercial and data protection dimension of AI in fashion

As well as delving the risk-based approached being adopted in the EU AI Act, an instructive discussion was the attribution of relevant powers arising from the act. Some Member States may seek to consolidate powers within existing organisations whereas others may use this as an opportunity to create newly organised authorities.

But there is good news for brands as preparation and compliance will likely hold (some may say hauntingly) close similarity to the preparatory works for GDPR and mapping and understanding the use of AI will likely be key. But the association with data protection doesn’t end there! Given the central importance of the use of data to modelling and implementation of AI, many aspects of the introduction of the technology will overlap with existing data protection obligations.

All things considered, the fashion industry will likely need to keep a broad purview as to what needs and demands are being placed upon it by the introduction of AI. In this regard, quick adaptation will be needed within the fashion industry. This may leave rightsholders concluding, however, that this is just business as usual.