Four things brands should consider when developing AI protocols

This article on AI ethics is part of our Vogue Business membership package. To enjoy unlimited access to our weekly Technology Edit, which contains Member-only reporting and analysis and our NFT Tracker, sign up for membership here. Fashion and beauty brands are increasingly experimenting with artificial intelligence to decrease time spent on menial tasks, enhance creative outputs and improve the customer experience. But the tech’s potential is clouded with concerns over job security and misinformation. A proposed AI bill in California — which was rejected by Governor Gavin Newsom in September — highlighted the tension between regulating generative AI and fostering innovation, and the limits and responsibilities of governments in that process. This is the latest in a series of high-profile conversations between technology companies and governing bodies to make sense of a tech that has the ability to mimic humans and act seemingly autonomously. The Biden administration issued an executive order concerning AI last November, while the European Union AI Act — considered the first comprehensive set of regulations for the development and use of AI — was approved in March this year. These regulations and proposals largely concern big tech, though their impacts trickle down to fashion, beauty and retail, helping shape consumer norms and providing guardrails for risk-averse brands. However, legislators and tech companies are still divided on how to regulate AI without stifling innovation. Plus, any law will take months to go into effect and is likely to be lacking in the types of industry-specific details that could comfort the fashion industry. Fashion and beauty brands should explore self-regulation, experts say. “Fashion, beauty and luxury businesses need to take their brand reputations really seriously. No doubt AI will help many of them in many ways, but a big mistake with AI could also do a lot of damage,” says Brian Green, director of technology ethics at the Markkula Centre for Applied Ethics at California’s Santa Clara University. “The more they do now, the more likely they are to protect their brand and be prepared for when more stringent regulations might come.” Some have already gotten started. Tapestry Group has an AI Centre of Excellence and a governance team to promote AI education, recommend and enforce policies, track AI developments, and foster cross-functional collaboration, says Tapestry chief information officer Yang Lu. It addresses data privacy, bias mitigation, regulatory compliance and ensures that Tapestry’s use of AI aligns with the organisation’s values, she adds. Hugo Boss has developed comprehensive internal generative AI guidelines to define the conditions to use the tech within the group (covering compliance with company policies and legal requirements), plus AI principles designed for employees to be aware of the limitations of AI and an internal “knowledge hub”. LVMH, meanwhile, has an AI Factory dedicated to establishing and implementing the group’s AI strategy and education. L’Oréal Group is establishing the uses of what it calls “intentional AI”, while the Estée Lauder Companies has an AI Innovation Lab and has internal guidelines and principles for generative AI across the company. Brands have an ethical and reputational impetus to proactively establish AI ethics protocols and best practices, says Vicky Brown, general counsel and commercial and chief privacy officer at advertising agency WPP. “To operate in this space in a transparent and ethically sound way, and to encourage consumers to want to purchase your brand, think about AI and your business through a practical lens, and the brand values that you have had in some cases for hundreds of years. The first principle is not, ‘Can we do this?’, but, ‘Should we do this?’ and, ‘Does it feel right?’ The answers to that are often extremely clear.” Here are four key areas to consider when putting AI protocols in place. Disclosure and transparency California regulations, set to go into effect in 2026, require developers to disclose AI system training data and when content has been AI generated. Disclosure and transparency have emerged as key themes among consumers as well; TikTok, YouTube and Instagram require people to label AI-generated content or automatically detect if content seems to have been generated by AI. Google’s AI-generated search results now include sources for its information. Many brands have found that their customers care more about if they were informed in advance than if creative output was impacted by AI. For example, a recent collaboration between Collina Strada and Baggu revealed that customers were not pleased when they learnt that some of the prints used in a series of bags had been created, in part, using generative AI (the product description pages disclosed that some had been “AI-conceptualised”). Eponymous designer Norma Kamali has been an enthusiastic adopter of AI to inspire designs, and has created a model (with consultancy Maison Meta) whose inputs are limited to Kamali’s 57-year catalogue of work. Even then, customers want to know. “You can get criticised for something that is AI, even if it’s all my intellectual property,” she says, pointing out some comments that the brand has received on social media, including some misguided assumptions that it is “using AI to sell clothes”. “Transparency for the consumer is essential,” Brown says. She recommends an icon to indicate an AI-generated image and its sources. This disclosure extends to when customers are engaging with an AI chatbot, or when and how a brand is using customer data, says Gina Bibby, head of the global fashion tech practice at law firm Withers. Some brands might elect to publicly disclose their AI principles. “Getting the discussion out there in plain English is a good thing, because there is a huge education piece for society, and it’s raising the bar for people’s understanding,” Brown says. Human likenesses AI tools enable brands to create human imagery, often to save time and money and to increase the types of models pictured in the clothing. This, too, can be hard to navigate. Levi’s generated considerable debate last year when it worked with tech company Lalaland to add AI-generated models to its e-commerce product imagery. L’Oréal Group has a policy of avoiding AI-generated life-like imagery, extending to faces, bodies, skin and hair, in external communications. “L’Oréal took a stance very early on that we wouldn’t generate human-like faces or skin to demonstrate the benefits of our products,” said Laetitia Raoust, chief digital and marketing officer for L’Oréal UK & Ireland, during a recent roundtable organised by WPP. “It doesn’t mean we won’t use generative Al — we’re very excited by generative AI — we use it for a lot of things, but we still think that beauty is very human.” “Full disclosure of when AI is being used is a good idea, especially if artificial pictures could create unrealistic beauty standards. But even better would be avoiding using AI-created human likenesses in the first place,” Green of Santa Clara University says. “Companies should really keep the well-being of their customers in mind, and the well-being of society. Choices like these decide what kind of world we are all going to live in — either one full of fake AI or real human beings. Knowing the right boundary between these two futures will be extraordinarily important.” Because beauty brands could uniquely use AI to augment human imagery and misrepresent the impact of products, they “have a unique opportunity to get ahead of regulation by regulating themselves”, says Brown. (She notes that there are already laws that prevent misrepresentative advertising.) Kamali doesn’t generate human-like models in editorial photoshoots because she likes the energy a real human brings. “If a model likes the dress I put her in, something else will happen. An avatar image doesn’t do that for me,” she says. Some brands are using AI to generate scenes, rather than humans, which anecdotally seems to be less polarising to consumers; Puma has experimented with tailored background imagery for certain regions, for example, meaning that a customer in Japan might see a shoe worn while hiking Mount Fuji. AI-created avatars are subject to existing regulations, pointed out Arent Fox Schiff partner Dan Jasnow, who co-leads the law firm’s AI, metaverse and blockchain group, during a recent symposium, meaning that disclosure requirements for sponsored ads and brand ambassadors still apply. Customer-facing tools Generative AI tools span in-house and customer-facing uses, ranging from chatbots and virtual sales assistants to internal data dashboards or tools that translate or modify marketing copy. In general, there is a higher level of scrutiny for customer-facing uses, Brown says. Some companies prioritise internal tools that aid employees, while remaining conservative in terms of customer-facing applications. LVMH has made the decision to avoid customer-facing uses of generative AI, although it does use AI to help make suggestions to sales associates. And for its recent innovation prizes, it recognised companies that generate marketing videos and lifelike jewellery imagery. Lingerie brand Adore Me has a tool that enables customers to submit a text prompt to create a custom print for a bra and panty set, enabling customers to co-create their own pieces. The label’s SVP of strategy Ranjan Roy says it monitors for any potential inappropriate or copyright-infringing content through a two-part system. If anything is flagged as concerning via an automated system, Adore Me’s in-person team reviews the content before it is sent to production. When it comes to chatbots and other tools that are created with third-party systems, experts advise that brands should review which training data their tech partners use, considering factors such as inclusivity and appropriate use of intellectual property. Even in-house uses of chatbots might be subject to a brand’s AI protocols. Some companies advise employees to avoid inputting proprietary documents or customer data into open systems such as OpenAI, as this could disclose private information. “Don’t just assume an AI tool is fit for use, as many tools are built on third parties; for example, many are using Chat GPT as the underlying large language model (LLM), so how is your vendor sharing data back with OpenAI?” Jasnow says. In this case, a company might want to enter into an enterprise agreement, meaning the data is contained within the company’s own system. Intellectual property and generated designs Many brands, including Balmain, Collina Strada, Coperni, Vaquera, Nike and Norma Kamali, have tested creating styles using generative AI. Experts advise that designers be aware of how their tools were trained, and the content they are putting into the models. Brands should be mindful of potential copyright infringement claims when using data sets that contain third-party materials, Bibby says. “You want to make sure creators are getting due recompense and being acknowledged for their creativity,” Brown advises. “You want to know what creative sources have gone into those models.” Kamali’s system uses only her designs. To create this, she manually grouped her pieces by category. She is able to generate prompts with specific needs, such as exclusive items made for retailers. This approach is most useful for brands overseen by the same designer who has a consistent and recognisable aesthetic. “It’s very hard when brands change their look. This works best when the brand’s visual identity is very clear,” says Kamali. She likens it to when a new designer revisits the archives of an historic brand. “When Karl [Lagerfeld] started at Chanel, I thought, how will that work? But he went to the archives, and he did a great representation of Chanel as a result.” Kamali, who recently created an AI installation in New York, faces a lot of curiosity from peers in the industry. She firmly believes that AI will be helpful in extending creativity and potentially even helping protect intellectual property and manufacture apparel going forward. “My purpose in sharing my exploration is to say that AI will be even more influential on the destiny of the world, just as electricity was in the Industrial Revolution. We need to be a part of this, despite our fears and concerns.” Sign up to receive the Vogue Business newsletter for the latest luxury news and insights, plus exclusive membership discounts. Comments, questions or feedback? Email us at feedback@voguebusiness.com.

Make better business decisions

More technology stories from Vogue Business

Make better business decisions