The intellectual property saga: AI’s impact on trade secrets and trademarks | Part 2

The European Union (EU) has reached a historic provisional agreement in 2023 on the world’s first comprehensive set of rules to regulate AI, which will become law, once adopted by the EU Parliament and Council. The legislation, known as the AI Act, sets a new global benchmark for countries seeking to harness the potential benefits of AI, while trying to protect against possible risks of using it. While much of the attention was given to parts such as biometric categorisation and national security, among others, the AI Act will also give new guidance on AI and copyright. 

The AI Act’s approach regarding Copyright and Transparency in AI takes a nuanced stance, requiring transparency regarding training data without demanding exhaustive lists of copyrighted works. Instead, a summary of data collections suffices, easing the burden on AI providers. Nonetheless, uncertainties persist about the foundation model of providers’ obligations under copyright laws. While the AI Act stresses compliance with the existing regulations, including those in the Digital Single Market Directive, it yet raises concerns about applying EU rules to models globally, potentially fostering regulatory ambiguity for developers.

In one of the previous blogs,  the Digital Watch Observatory elucidated the relationship between AI-generated content and copyright. The analysis showed how traditional laws struggle to address AI-generated content, raising questions of ownership and authorship. Various global approaches – denying AI copyright, attributing it to humans – highlight the complexity.

This part will delve into the influence of AI on Intellectual Property Rights, and will assess the ramifications of AI on trade secrets and trademarks focusing on examples from the EU and US legal frameworks.

Trade Secrets and AI Algorithms

Within the realm of AI and intellectual property, trademarks and trade secrets present unique challenges and opportunities that require special attention in the evolving legal landscape. As AI systems often require extensive training datasets and proprietary algorithms, determining what constitutes a protectable trade secret becomes more complex. Companies must navigate how to safeguard their AI-related innovations, including the datasets used for training, without hindering the collaborative nature of the AI development. 

Trade secret laws may need refinement in order to address issues like reverse engineering of AI algorithms and the accidental disclosure of sensitive information by AI systems. However, given the limitations associated with patenting and copyrighting AI-related content, trade secret principles seem to present an alternative, at least in the USA. Patents necessitate a demonstrated utility disclosed in the application, while trade secrets lack this requirement. Trade secrets cover a broader range of information without the immediate need to disclose utility. In addition, trade secret law allows information created by an AI system to be protected, even if the creator is not an individual. This differs from patent law, which requires a human inventor listed on the application. 

Trade secrets, traditionally associated with formulae and confidential business information, now extend to AI algorithms and proprietary models. Safeguarding these trade secrets is critical for maintaining a competitive edge in industries in which AI plays a pivotal role. In the USA, trade secret law safeguards a broad spectrum of information, encompassing financial, business, scientific, technical, economic, or engineering data, as long as the owner has taken reasonable measures to maintain its secrecy, and the information derives value from not being widely known or easily accessible through legitimate means by others who could benefit from its disclosure or use (as defined in 18 U.S.C. §1839(3)). It is important, however, to consider that patent owners have a monopoly on the right to make, use, or sell the patented invention. In contrast, owners of AI-based trade secrets face the risk of competitors reverse engineering the trade secret, which is permitted under US trade secret law.

Requirements related to secrecy exclude trade secret protection for AI-generated outputs that are not confidential, such as those produced by systems like ChatGPT or Dall·E. Nevertheless, trade secret laws seem to be more flexible to safeguard various AI-related assets, including training data, AI software code, input parameters, and AI-generated content intended solely for internal and confidential purposes. Importantly, there is no stipulation that a trade secret must be originated by a human being, while AI-generated material is treated like any other form of information, as evident in 18 U.S.C. §1839(4), which defines trade secret ownership.

Instead of pursuing patents, based on traditional laws that seem to provide ambiguous guidance on AI and Copyright,  numerous AI innovators opt for trade secret protections to safeguard their AI advancements, as these innovations in commercial use frequently remain concealed and difficult for others to detect. With the AI Act soon to become law, there’s a likelihood that the EU will necessitate disclosing how AI innovations operate, categorising them as limited or high risk. This consequently leads to trade secret safeguarding to no longer be viable in some instances. 

Establishing clear guidelines for what qualifies as a trade secret in the AI domain, and defining the obligations of parties involved in AI collaborations will be essential for fostering innovation while ensuring the protection of valuable business assets.

Trademarks and Branding in the AI Era

The integration of AI technologies into product and service offerings has also reshaped the landscape of trademark protection, presenting both challenges and opportunities for businesses. Traditionally associated with logos, brand names, and distinctive symbols, trademarks now extend their scope to encompass AI-generated content, virtual personalities, and unique algorithms associated with a particular brand. As companies increasingly rely on AI for customer interactions, the challenge of maintaining brand consistency in automated, AI-powered engagements becomes paramount. In the realm of AI-driven customer service and chatbots, the traditional understanding of the ’average consumer’ in trademark infringement cases undergoes transformation. When an AI application acquires a product with minimal or no human involvement, determining who, or more crucially, what constitutes the average consumer, becomes a pertinent question. Likewise, identifying responsibility for a purchase that results in trademark infringement in such scenarios becomes complex.

While there have been no known cases directly addressing the issue of AI and liability in trademark infringement, there have been several cases within the past decade adjudicated by the Court of Justice of the European Union (CJEU) that could offer insights into the matter when considering this new technology. For instance, the Louis Vuitton vs Google France decision focused on keyword advertising and the automatic selection of keywords in Google’s AdWords system. It concluded that Google wouldn’t be accountable for trademark infringement unless it actively participated in the keyword advertising system. Similarly, the L’Oréal vs eBay case, which revolved around the sale of counterfeit goods on eBay’s online platform, determined that eBay wouldn’t be liable for trademark infringement unless it had clear awareness of the infringing activity. A comparable rationale was applied in the Coty vs Amazon case. 

It would seem that if a provider of AI applications implemented adequate takedown procedures and had no prior knowledge of infringing actions, they would likely not be held responsible for such infringements. However, when the AI provider plays a more active role in potential infringing actions, the two cases indicate that the AI provider could be held accountable. 

In the case of Cosmetic Warriors Ltd and Lush Ltd vs Amazon.co.uk Ltd and Amazon EU Sarl before the United Kingdom High Court, in 2014, Amazon was determined to be liable for trademark infringement. Amazon used ads on Google mentioning ’lush’ to bring people to its UK website, where Lush claimed Amazon was breaking trademark rules by showing ‘LUSH’ in ads and search results for similar products without saying Lush items weren’t available on Amazon. The Court explained that consumers were unable to discern whether the products being offered for sale were those of the brand owner or not, thus illustrating that the evolving definition of the average consumer and the delineation of responsibility in trademark infringement cases involving AI require nuanced legal considerations. 

Conclusion

As AI continues to impact various industries, the ongoing evolution of intellectual property laws will play a pivotal role in defining and safeguarding AI innovations, underscoring the need for adaptable regulations that balance innovation and protection. The intersection of AI and intellectual property introduces novel challenges and opportunities, necessitating a thoughtful and adaptive legal framework. One crucial aspect involves the recognition and protection of AI-generated innovations. Traditional IP laws, such as patents, copyrights, and trade secrets, were designed with human inventors in mind. However, the autonomous and generative nature of AI raises questions about the attribution of authorship and inventorship. Legal systems will need to address whether AI-generated creations should be eligible for patent or copyright protection and, if so, how to attribute ownership and responsibility. This demands a forward-thinking approach from policymakers, legal scholars, and industry stakeholders to craft a legal landscape that not only accommodates the transformative potential of AI, but also safeguards the rights, responsibilities, and interests of all parties involved.