1. The Al Boom
1.1 Hyperintelligence or just hype?

As with any fast-spreading innovation, the huge fanfare surrounding Al (and in particular ‘Generative Al’) has been met with intense scepticism, and many have cast doubts as to its capabilities, industry adoption, and ultimately its long-term staying power.

It is true that Al is still in relative infancy, and exhibits limitations. Al can generate incorrect answers, or even ‘hallucinate’ by seemingly making up information. Such hallucinations have made a presence in the law, with Steven Schwartz, a New York attorney admitting using ChatGPT for a case, by virtue of his brief containing completely fake cases and judicial opinions generated by an Al tool.

Despite such limitations and concern about Al’s true potential, its growth is unstoppable — the world’s third largest tech company and leader in Al chip production, NVIDIA, experienced a massive stock-surge, spurred by the growing confidence in Al as constituting the future.  According to Toshiya Hari for Goldman Sachs, NVIDIA’s stock is expected to surge another 22%, despite almost doubling in 2024 already.

The UAE was the first nation to appoint a Minister for Artificial Intelligence and has shown its eagerness to be involved early, by acquiring thousands of NVIDIA’s chips last year.

The UAE has developed its own Al capabilities without over-reliance on the US or China, and has recently launched the Falcon 2 series via its Technology Innovation Institute, further pushing UAE to the top of the global Al landscape. [5] In addition, in June 2024 the UAE launched a Dubai Universal Blueprint for Al to enhance the status of Dubai as an internationaltech hub. 22 Chief Al Officers were appointed across government entities in Dubai.

OpenAl revealed GPT-40, a free iteration of GPT-4 purported to be “faster..better at understanding images, and speaks more languages”. [6] The growing competitiveness between OpenAl and Google could soon match the infamous Microsoft and Apple rivalry.While much caution over Al’s potential for everyday application remains, the chances of an ‘Al winter’ happening anytime soon seem slim.

Rapid innovation continues to force our expectations of Al upwards on a seemingly daily basis. This, combined with continued mass investment,  suggests the peak is still far off — a both scary and exciting prospect. ‘Amara’s Law’, coined after the scientist and futurist Roy Amara, seems especially relevant here:

“We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run”.

Applying this to Al, there is healthy scepticism to overzealous claims by those with a vested interest in Al’s success, yet we acknowledge that the long term capabilities are not defined by its current teething phase. An example of this is Sam Altman’s recent claim that we could be one or two ‘ideas’ away from artificial general intelligence (AGI), i.e. theoretical Al systems that possess capabilities beyond that of humans.

1.2 Al’s application to the fashion industry

Creative industries like fashion are particularly susceptible to the influence of Generative Al, in terms of visual and audio content. Industry experts have widely discussed how Al has and will continue to affect the world of fashion, with McKinsey & Company estimating that Generative Al could in the next 3-5 years, “add $150bn, conservatively, and up to $275bn to the apparel, fashion and luxury sectors’ operating profits”.

McKinsey states the potential for media generation as one major possible driving force of Al- fuelled creativity in fashion, such as for the use of marketing campaigns.

The Business of Fashion, a global platform that provides insight into the industry, expanded on the analysis and identified 4 opportunities Generative Al offers to fashion:

1. Enhance and support creativity;

2. Augment and hyper-personalise customer-facing content and services;

3. Reduce costs or complexity of ambitious projects; and

4. Automate manual tasks.

Much of the observations on the use of Al in other industries may be applied to fashion, such as the (current) need to add considerable human touches to Al-generated content, and that Al simply isn’t and will never be a replacement for creativity. Although a limitation on the surface, those within creative industries may view this as a sweet point in keeping Al as a useful tool to serve creatives rather than replace them.

Despite not being as susceptible to ‘Al-replacement’ as other industries by account of its inherent need for human creativity, large fashion companies are not likely to shy away from using Al to the expense of humans if it helps business.

Such examples are evident with brands like Louis Vuitton and Nike partnering with Al modelling companies that may not only threaten models, but also other industry professionals involved in photoshoots. Levi’s in particular garnered backlash after stating they would use Al models to promote diversity, without hiring actual models from diverse backgrounds .

An area outside design, where Al may likely have a less threatening impact, is in customer experience enhancement. An example of this is the ChatGPT-operated tool ‘Madeline’ launched by Kering, the second biggest luxury goods company behind LVMH, hosting brands like Gucci, Balenciaga, Alexander McQueen, Brioni and more.

Critics have stated that despite initial excitement surrounding Madeline and other chatbots in the space, they have not been able to revolutionise the industry in the way that may have been expected. Like many prior automated chatbots, reviewers felt that the new ChatGPT-powered chatbots still lacked the utility that a human assistant offers, and often returned incorrect or irrelevant responses to queries from users.[14]

OpenAl have unveiled an upcoming video generation software, entitled ‘Sora’. [15] In OpenAl’s own words, ‘Sora is an Al model that can create realistic and imaginative scenes from text instructions’. OpenAl has showcased multiple videos created by Sora, which from initial viewings are simultaneously breathtaking and terrifying in living up to OpenAl’s touting of the tool’s realism.

Discussing what this means for fashion, The Fashion Law (‘TFL’) mentions its capacity for marketing & advertising, including greatly enhancing the customer experience outside physical stores.[16] The accuracy in terms of the way fabrics look and behave in certain Sora-generated videos could allow brands to kill two birds with one stone — firstly, by being able to massively cut visual marketing costs; and secondly, being capable of producing videos that may have previously only been possible on a blockbuster budget, while maintaining clothing detail that previous Al iterations have lacked (eg DALL-E or Midjourney).

Existing fashion tech have begun to incorporate Al into their tools, such as the garment design and simulation software, CLO. In their latest update, they have revealed an ‘Al Texture Generator’, allowing users to generate fabric textures in seconds via textual prompts.

2. Al & the Law: achieving balance between

innovation and regulation

While much of the discussion on Al’s impact on fashion has centred on exciting possibilities in how it can assist design and marketing, this is naturally accompanied by discussion on the uncertainty surrounding protective regulations, particularly pertaining to intellectual property matters, and how such regulation will balance protection with innovation.

2.1 Evolving global Al legislation

The novelty and rapidly changing landscape of Al has forced governments and authorities to attempt to keep up by formulating official approaches as to how they intend to regulate the space, with over 30 countries passing Al laws to date, as well as 75 national Al strategies being released to date and many more currently in development.[18] OECD’s Al Policy Observatory offers a live bank of over 1000 Al policy initiatives, where strategies and governance can be tracked and analysed.[19]

EU

Constituting the world’s first comprehensive legislation on Al is the European Union’s Al Act, passing through parliament on March 2024.[20] The act aims to prioritise safety by classifying Al systems based on risk, with those that fall under ‘unacceptable risk’ being prohibited by Article 5 (consisting mainly of systems whose applications infringe on citizen’s rights, manipulate human behaviour and exploit human vulnerabilities).[21]

Other classifications include high risk; limited risk; and low or minimal risk, with requirements naturally loosening when risk lowers.

This legislation will likely have a much wider effect in how Al legislation is approached by other nations and authorities, resulting from both its landmark nature and the EU’s reputation of influence.

The biggest question from the EU Al Act is whether it can avoid stifling innovation while still maintaining safety. France had previously been opposed to the legislation for this reason, as it felt the Al Act could hamper the global competitiveness of its startups, before eventually agreeing to ratify the Act after negotiations to provide some protections for innovation in balancing technological transparency with business secrecy.[22] Some answers to this question may be beginning to surface, with reports that European startups are being approached by the UAE to relocate their operations with the enticement of financial incentives and looser regulations.[23]

UK

The UK has generally differed from the EU by adopting a ‘pro-innovation’ regulatory stance towards Al which it outlined in its March 2023 white-paper.[24] Rather than enacting Al-specific legislation, the UK proposed a framework with principles for existing regulators to interpret in line with their specific sectors. Following consultation, the government issued a response in February 2024, in which it reiterated its willingness to stick to a pro-innovation approach without specific Al legislation. It however included that as well as being ‘pro-innovation’, it is also ‘pro-safety’, and acknowledged calls for further clarity on how existing laws will apply to Al, as well as admitting that at some point heavier regulation will be required, but it is not confident that this is currently necessary .[25]

In the short time since then, it has surfaced that the government may have started to draft new legislation which will make sharing LLM (large language models e.g. OpenAl’s GPT or Google’s Gemini systems) algorithms and evidence of safety testing with the government mandatory. In late 2023, a private members bill called the ‘Artificial Intelligence (Regulation) Bill’ was proposed, which aims to establish an ‘Al Authority’ to ensure a more consistent, hands-on approach to Al across sectors.[26]

USA

There is currently no direct Al-specific legislation at a federal level in the US, although some existing laws do cover Al at a limited level. Some regulatory frameworks have been introduced, such as the ‘White House Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence’[27] States have begun to enact their own legislation, such as the ‘Utah Artificial Intelligence Policy Act’, making it the first state to pass Al-specific legislation on consumer protection.

UAE

The UAE’s own national strategy (‘UAE National Strategy for Artificial Intelligence 2031’) was published in 2018, with cabinet approval in 2019. [29] [80] It contains 8 objectives centred on making the UAE global Al hub, as well as “ensurling] strong governance and effective regulation”. So far, the UAE has adopted a ‘soft’ approach to regulation by favouring general guidelines instead of legislation, overseen by its Al and Blockchain Council.

A balanced approach is favoured by the UAE, which seems to align with sentiments echoed by many Al pioneers that we should focus on facilitating innovation now when Al is in general infancy, and regulate more heavily when the technology starts to near the risks that reaching AGI will inevitably carry.

The UAE’s Minister of State for Artificial 2.2.1.1 Copyrights Intelligence, His Excellency Omar Sultan Al Despite the UK’s approach in handing the reins to Olama, has backed Altman’s statement that the sector regulators, UKIPO had earlier failed to reach UAE could act as a “regulatory sandbox” in an agreement between Al companies and rights testing Al and devising suitable regulation that holders within the creative industries to establish a could impact global Al regulation.[31] voluntary code of practice.[33] There are reports of the UK considering a shift from this approach The UAE is not merely ‘pro-innovationr’ like the towards introducing some legislation, possible UK, but is an active player in the Al space, here to amendments of copyright law could take place to rise to the top rather than to merely take part. allow entities to opt-out from their data being used by LLM’s to train themselves.

2.2 How Al is affecting fashion law

When looking at how Al’s legal and regulatory landscape will translate to fashion law, various individual practice areas that make up fashion law are important to understand: intellectual property, data protection & privacy, and employment law.

In general, tougher regulation like the EU Al Act may be viewed as more favourable to the industry, in creating clarity and protections on how Al can be used for designing, marketing, consumers and e- commerce.

Lawyers advising the fashion industry will be required to religiously keep up to date with changing Al technology and regulation, so that they can best advise fashion companies on how they can simultaneously take full advantage of Al’s capabilities while also mitigating risks by being aware of limitations and complying § with regulations. Businesses should act fast by taking lessons from the GDPR in avoiding under- preparation that marred businesses at the time.

2.2.1 Intellectual Property Law

As Al continues to grow, so have IP cases against Al companies, usually involving infringement claims concerning the use of copyrighted material by Al systems to train themselves. Famous ongoing cases include New York Times v OpenAl and Microsoft, where the defendants argue any infringement may be justified by ‘fair use’ exceptions. [32]

Those in the fashion industry may not be satisfied by this, and will likely push for an opt-in rather than opt-out arrangement, as proposed by FT CCO Jon Slade to the House of Lords Digital and //\ Communications Committee.

SOL

The novel question of whether solely Al- generated works may be copyrighted, and therefore protected, has been answered in the negative by the US Copyright Office (USCO), with exceptions for works containing, “sufficient human authorship”.[37] In the UK they are widely seen as falling into the definition of computer- generated works, as per $178 of the Copyright, Designs and Patent Act 1988 (CDPA), although this is still unclear pending future clarification or amendments to the law.[38]

The EU Al Act does not address this issue and currently seems in line with the US with its existing laws. The UAE shares some similarities with the UK in widening protected works in 2021 to include, “smart applications, software and software applications, databases and similar Works…”.[39] The definition of an ‘author’ leaves room for interpretation that could potentially include solely Al-produced works, but again requires further clarification.[40]

For fashion companies, restrictive rulings may mean solely Al-generated designs cannot receive copyright protection. A practical example being where a brand uses Al to generate an aspect of an overall piece, e.g. an embroidery design for a bag. The parts of the end product that involved ‘sufficient human authorship’ may be eligible for copyright protection, but if the embroidery design was solely generated by Al and not subject to human alteration, it would not be protected.

2.2.1.2 Patents

Although typically less touched on in fashion law compared to copyright, the growing innovation involving the use of tech in fashion products, termed ‘wearables’ or ‘smart-textiles’, has raised the importance of patent laws in_ their application to fashion. Textile innovation has included well-known fabrics like Lycra and Gore- Tex (the latter of whom’s main patent has since expired).

As textile innovation has developed, the use of technology in the development of smart textiles has continued to grow, with the respective market being valued at $878.9m, and an expected CAGR (compound annual growth rate) of 31.29% over 2021 to 2025.

Examples of innovation include fabrics which monitor the wearer’s health, from heart rate to body temperature to even ensuring proper posture.

The biggest story involving Al’s impact on patents is whether Al can legally be called an ‘inventor’. In the UK, the Supreme Court ruled that Al could not be named as an ‘inventor’ in a patent application and requires a human, which was similarly upheld by US federal courts and EU courts.

In the UK, the ruling did not allow the human who owned the system to gain ownership of inventions solely generated by Al, but stated that an applicant who used Al (i.e. with human oversight) could be entitled to a patent. The Supreme Court admitted that the 1977 UK Patents Act would have to be amended if patents were to be granted to wholly Al-generated inventions.

In the context of fashion, this may further incentivise human input in using Al to develop products like wearables. At the moment, this is unlikely to be a major issue given that Al has not yet achieved the ability to create without human oversight, but as Al continues to rapidly develop and full automation becomes viable, further legal clarification will be required.

The UAE has not currently addressed the question, but given the relative uniformity in rulings by other leaders in Al legislation, it is likely to adopt a similar approach, especially given current similarities with IP laws in the UK.

The UK Court of Appeal is considering a case determining the patentability of artificial neural networks (ANN).[48] ANN systems have been used in fashion (e.g. Tommy Hilfiger partnering

with IBM and The Fashion Institute of Technology) to optimise and _ personalise customer experiences, forecast trends and

influence designs.The outcome of this case will shine light on how Al will continue to be utilised in fashion. ;

2.2.2 Data Protection & Privacy Law

Just as Al presents novel and exciting ways in which brands may interact with consumers to enhance customer experience, it also comes with important legal considerations in protecting consumer rights. Although laws such as the GDPR cover these issues to an extent, they do not properly regulate the use of Al.

The EU’s Al Act ‘high-risk’ category will likely catch software such as virtual styling/fitting rooms, as they need biometric data like physical measurements to operate. Less likely is the possibility that other software involving biometric data falls into the banned ‘unacceptable risk’ category — it is nonetheless important for both developers and brands that utilise their technology to comprehensively examine its use of Al to avoid hefty penalties associated with non-compliance.

Although developers will have more stringent obligations to follow than ‘deployers’ (entities that utilise Al systems for professional activities e.g. fashion companies), the latter are still bound by strict requirements, including monitoring the operation of high-risk Al systems, assigning human oversight, and keeping the logs generated by these systems for an appropriate period of at least six months generally.

According to Vogue Business, fashion-tech developers of such systems are beginning to prepare for such regulations, with the Al styling platform Intelistyle enabling ‘deployers’ of its tech to control how its system uses consumer data, in order to, “manage legal risk without slowing down Al adoption”.[51] The Vogue report reiterates notions that the EU Al Act will set a precedent in the same way the GDPR did, and that the industry should start as soon as possible in preparing how they intend to deal with regulation compliance.

Image right issues have surfaced in relation to the use of Al models, with cases of real models having their likeness used and altered to produce Al fashion models.

Alongside IP and ethical considerations, this also raises data privacy concerns, and will require companies to disclose details around any use of Al models. Such provisions were included in an amendment to New York’s proposed Fashion Workers Act, requiring. “clear, conspicuous and separate written consent” for the creation or use of a model’s ‘digital replica’, along with requiring certain details as to how they are used.

In the UAE, the DIFC has enacted regulation pertaining to data protection, with Regulation 10 requiring deployers of Al systems to give a detailed notice to users of the processing of personal data by systems that lack human oversight (called ‘autonomous and_ semi- autonomous’, including Al). In addition to this, systems should abide by some general principles to be ethical, fair, transparent, secure and accountable.

2.2.3 Employment Law

Al may assist in making employee management more efficient, such as speeding up recruitment processes and reducing human bias, as well as better appealing to potential talent, The application of Al for recruitment has seen use in the fashion industry, such as Kim Kardashian’s ‘Skims’ brand using Dweet Al to create job postings and subsequent shortlists.[55]

With the EU Al Act’s categorisation of Al systems by risk, the application of Al in employment practices could in many cases be labelled ‘high- risk’, such as its use for recruitment and subsequent management of employees.[56] In addition to the requirements laid out in the above section, fashion companies will be required to inform affected employees and employee representatives of any introduction of high-risk Al systems into the workplace.[57]

The proposed ‘No Robot Bosses Act’ in the US would serve to protect current and prospective employees by banning their employers from exclusively using Al without human input for decisions such as hiring and firing. This would also involve such systems having to undergo testing to ensure they do _ not_ involve discrimination, as well as employers having to be transparent about when and how they are being used.[58] An infamous example of this was Amazon discontinuing an Al recruiting system which exhibited clear male bias, which it had taught itself over a ten year period.[59]

Conclusion

Whilst fashion law has always featured technology-adjacent areas such as data protection & privacy and IP, the influence that Al has and will have on the industry, including designs, marketing and customer experience, will continue the intertwining of fashion and technology both in the opportunities and legal challenges both face. It remains to be seen how regulation and legislation will handle Al when the serious risks associated with superintelligence inch closer to reality, marking this area of law as uniquely intriguing in its novelty and development pace, as well as its international relevance and impact. The fashion industry will undergo a tsunami of change with Al as will most industries, for better or for worse.