AI and Fashion: Between creativity and innovation
Recently, retail giants have embraced an algorithmic approach to fashion. From apps that provide feedback or advice on outfits to interactive fitting rooms with mirrors that recognize the clothes you are wearing and suggest other matching options based on style, color and mood, AI seems to be the latest trend in the fashion industry.
The creation of AI-based fashion items takes the relationship between innovation and creativity to a new level and raises a series of questions, some of which are still unanswered. Who is the owner of the creations made by AI? Can AI infringe upon the intellectual property rights of others? What kind of personal data is collected by AI? Is the resulting data processing subject to data protection regulations? Will AI enhance or diminish creativity in the fashion industry?
Copyright protection for works created by AI
Under Italian law, creative works must be original to obtain copyright protection, and traditionally the requirement of originality has been linked to the natural person of the author. In fact, according to Article 6 of Law No. 633/1941 (Copyright Law), “the original legitimacy for the acquisition of copyright consists in the creation of the work, as a particular expression of the intellectual work” of the author. Therefore, machines and AI seem to be excluded from the notion of authorship.
This position was reiterated in a very recent decision by the US Copyright Office, which denied registration of AI-generated images on the grounds that they were considered unprotectable under US copyright law. The Office noted that registration of an original work is only possible if it was created by a human being since copyright law protects only “the fruits of intellectual labor” that are “based on the creative powers of the mind”, and such discipline is limited to the protection of the author’s “original intellectual conceptions”.
In accordance with case-law (Feist Publ’ns, Inc. v. Rural Tel. Serv.; Burrow-Giles Lithographic Co. v. Sarony) and legislation (see 17 U.S.C. § 102(a) and (b)) on copyright law, the US Copyright Office held that copyright protection cannot be granted to works made by a machine or mechanical process that operates randomly or automatically without sufficient creative input or intervention by a human author.
Similar conclusions were also reached at the national level with a recent Italian Supreme Court ruling on the recognition of copyright protection on a digital work depicting a flower, created through the use of software, that was used in a past edition of the Sanremo Festival.
In 2016, RAI used a digital work depicting a flower found on the web as a set design for the famous song festival without seeking permission from the author, who had created it some time earlier through the use of software. For this reason, two years later, the author sued RAI before the Court of Genoa, claiming infringement of her copyright on the work.
Confirming the decision of the judges at first instance and on appeal, the Supreme Court dwelt on the legal concept of creativity, holding that it does not coincide with that of creation, originality and absolute novelty. According to this interpretation, the creativity protected by Italian law is given by the personal and individual expression of an objective content belonging to the categories listed in Article 1 of Copyright Law. Therefore, an intellectual work can be protected provided that a creative act, albeit minimal, susceptible of manifestation in the outside world can be found in it.
In the case at hand, the digital image used in the set design of the Sanremo Festival did not simply represent a reproduction of a flower, but was a true reworking, deserving of protection under copyright law, since it was an original and creative idea coming from its author. Therefore, the Supreme Court rejected RAI’s defense that the author’s process was limited to choosing an algorithm and approving the result generated by the software, pointing out that the use of software does not exclude the elaboration of an intellectual work protectable by copyright, but only requires scrutinizing its rate of creativity more rigorously.
Therefore, algorithmic works of art are eligible for copyright protection as long as human choices are involved. In the case of an entirely AI-designed fashion collection, the person who developed the software could enjoy copyright protection for both the software itself and the work created through its use, and presumably be held liable in the event of infringement of third parties’ intellectual property rights.
AI in big data analysis
In such data-driven system, fashion brands are increasingly investing in AI to offer their customers innovative products and services. In fact, fashion houses are not only focusing on the use of algorithmic works that we have just analyzed, but they are also reinventing their websites and e-commerce channels by hiring virtual sales assistants, e-concierges, and chatbots capable of interacting with customers to help them choose and try out their favorite outfits, even from the comfort of their homes, while providing them with a new and exclusive shopping experience.
In this new reality, a crucial role is undoubtedly played by the data collected through websites, online applications, and social networks, which allow brands to understand which clothing items are trending and predict new trends. The fuel of the fashion industry can be found precisely in the use of big data and predictive AI systems. When AI is integrated into digital properties, brands can profile their customers and users to understand their interests and preferences, create targeted marketing campaigns, and provide personalized services based on their needs.
However, targeting users and customers of brands through big data analysis and the use of algorithms is considered processing of personal data. For this reason, fashion houses that decide to operate online and offer innovative services to their customers must comply with the requirements provided by data protection regulations, particularly with the obligations and responsibilities established by the GDPR. For example, fashion brands must collect and process only the personal data that is strictly necessary for as long as it is needed, in compliance with the principle of minimization set out in Article 5 of the GDPR.
Moreover, to lawfully collect and process personal data, brands have to identify the appropriate legal basis, which, in the case of optimized profiling using AI systems, must be obtained through the consent of the data subject. The GDPR also introduces specific information and transparency obligations: Article 13 of the GDPR requires the data controller to provide information about the processing activities, its purposes and methods, the existence of automated decisionmaking (including profiling), the underlying logic, and the potential consequences for the data subjects.
Lastly, the performance of a Data Protection Impact Assessment (DPIA) according to Article 35 of the GDPR is also crucial, as this type of processing can be invasive and entail a high risk to the rights and freedoms of the data subjects.
Monitoring and profiling activities of online users through algorithms and other big data analysis tools can also be optimized by combining data collected directly from users with those inferred from their behavior (eg custom audience matching and look-alike activities), including through cookies and other tracking technologies. These mechanisms are not only used to monitor users on the web but also to influence their behavior and choices, particularly regarding consumers’ purchasing decisions, sometimes undermining individual autonomy and freedom.
Brands must take specific precautions when intending to adopt user tracking mechanisms. The Italian Data Protection Authority (Garante per la protezione dei dati personali) has also intervened on this point. With its Guidelines on cookies and other tracking tools issued on June 10, 2021, it reiterated the circumstances in which the installation and use of cookies (or other tracking tools) can occur on users’ devices and the rules for analyzing data collected through such technologies. In particular, to lawfully install cookies or other tracking technologies, brands operating online must: (i) inform users with a short privacy information notice (the cookie banner) that redirects to an extended privacy information notice (the Cookie Policy) outlining the purposes and methods of data processing through these technologies, and (ii) obtain a specific consent, which means a clear positive action from the user consciously consenting to the installation of cookies.
When AI becomes real, CGI is born
Thanks to the use of AI, brands can not only profile users and analyze large amounts of data, but they can also offer users renewed and innovative shopping experiences. In fact, AI enables the generation – and bringing to life – of different real-world personalities in the virtual world. These personalities are digital robots that simulate hyper-realistic human appearances, qualities, expressions, behaviors, and characteristics, going beyond common stereotypes.
A brilliant example of this is virtual influencers, known as CGI (computer-generated imagery), who, in their virtual form, promote renowned brands, endorse their products and services, and act as ambassadors for principles and rights.
Virtual influencers are the latest trend on social media. Currently, social networks host over 200 virtual influencers, with the most followed ones including Nefele, the first Italian virtual influencer and advocate for inclusion and diversity values, the twins Eli and Sofi, Shudu, Lil Miquela, Noonoouri, Rozy, and the popular virtual streamer CodeMiko on Twitch. Many fashion brands are already turning to these CGI influencers for promoting their products. For instance, Rihanna Shudu to promote her cosmetics brand Fenty Beauty, while Prada, Chanel, and Fendi have been collaborating with the famous Lil Miquela for years.
This demonstrates that robots can also set new fashion trends. Sometimes, collaborating with virtual influencers rather than real people even brings advantages to fashion brands. In the real world, there is always an element of uncertainty due to human free will. In the virtual world, brands can exercise complete control over influencers, determining how they should present themselves to their followers both aesthetically and communicatively. This control helps minimize reputational damages resulting from behaviors or initiatives unrelated to brand endorsement.
Even though they are virtual, influencers still need to comply with the regulations of the real world. For this reason, transparency obligations in advertising also apply to these new forms of virtual life. Furthermore, it is necessary to explicitly disclose the non-real identity of these characters on the platforms through which they convey advertising messages. Users should be aware that they are interacting with an avatar and not a human. This idea stems from the belief that the development of reliable technologies, both ethically and morally, should serve as useful tools for society and be aligned with human needs.
Moreover, it’s interesting to note how avatars and AI systems can be linked to forms of “abuse” of images, as in the case of deep fakes. This can result in more serious violations of personality rights, such as potential damage to one’s image, reputation, and digital identity, non-consensual dissemination of private images, potential risks of phishing and vishing attacks, identity theft, and breaches of security measures based on biometric recognition.
In this context, it’s essential that the systems used are cyber resilient, and the importance of adopting incident response schemes to remedy negative impacts on the business and protect the virtual influencer’s and brand’s reputation becomes apparent. This includes developing a post-attack remediation plan to mitigate potential claims for damages or other forms of liability.
It’s evident that the development of these new technologies requires a regulatory intervention to prevent abuses. Meta, for example, has started working with its developers and experts on an Ethical Framework to establish clearer boundaries for the use of avatars and virtual influencers, considering both potential harms and benefits. Currently, Meta’s primary aim is transparency and the importance of always providing the possibility to distinguish between what is real and what is not.
