Last week saw news emerge of a lawsuit filed by Apple against OpenAI for the alleged misappropriation of trade secrets. Apple claims that OpenAI selectively recruited former Apple employees, encouraging them to disclose and explain Apple’s proprietary information. According to the allegations, some of these former employees even downloaded confidential Apple information and took it with them when they left the company. This is by no means unusual. Trade secrets have always existed, but over roughly the last decade companies have increasingly relied on trade secret protection as a means of safeguarding their intangible assets. There are many reasons for this: the flexibility of the legal framework, the wide variety of assets it can protect, its prominence in sectors characterised by short innovation and obsolescence cycles, and the growing use of generative AI as a research tool, which often produces outputs that may not be protectable under other categories of intellectual property.
The AI sector in particular is especially conducive to the use of trade secrets as a mechanism for controlling technological advances. Examples include the datasets required to train and test systems, use-case definitions and business rules, training methodologies, refined algorithms, models, parameters, hyperparameters (so important for optimising accuracy and efficiency), behavioural rules governing systems, security measures, and much more.As in any highly competitive sector with several major players, it is only natural that intellectual property disputes should arise sooner or later. Given the growing importance of trade secrets, it is equally natural that they have become the weapon of choice in such disputes.
A few months ago, xAI, Elon Musk’s company and the developer of Grok, also brought a trade secret misappropriation claim against OpenAI, although the case was ultimately dismissed by the court. xAI alleged that OpenAI had recruited its employees with the intention of obtaining xAI’s confidential information. It further claimed that one former employee had downloaded the source code of the Grok model in order to provide it to OpenAI. Source code is, of course, a highly valuable repository for many of the AI assets mentioned above. In this respect, the claims brought by xAI and Apple are remarkably similar: targeted recruitment of employees and the alleged theft of information by those individuals. California does not permit post-employment non-compete clauses, which explains part of the issue.
However, such clauses are enforceable in Spain and in many other jurisdictions, including certain US states such as Massachusetts, home to the major pharmaceutical and biotechnology innovation hub centred around Boston. Post-employment non-compete clauses are not designed to prevent the misappropriation of trade secrets. Rather, they seek to prevent employees from placing their professional expertise at the service of competitors, something that would otherwise be entirely lawful in the absence of such restrictions. If an employee removes confidential information from a previous employer and uses it for their own benefit or that of a third party, that conduct is unlawful regardless of any non-compete agreement.
Apple’s claim nevertheless departs from the familiar narrative within the AI sector because it is not really about AI itself. The alleged misappropriation concerns everything needed to give AI a physical form: to make it tangible and place it in the user’s pocket.
OpenAI does not want to remain dependent on third-party platforms, such as Apple’s iPhone, one of the most successful consumer devices ever created. Its strategic ambition is therefore to develop its own line of hardware devices. This would provide a source of revenue, but more importantly it reflects a broader vision of a new paradigm in human-AI interaction. OpenAI aspires to achieve for AI hardware what Apple achieved with the iPhone: capturing the commercial rewards of defining a new technological standard while establishing the framework through which users engage with the technology. In a market where the capabilities of leading AI models increasingly converge, such a shift could prove decisive in attracting and retaining a substantial customer base.
However, moving from software development to the manufacture of mass-market consumer devices is not something that can simply be learned overnight. It requires enormous investment, extensive time, specialised expertise and scarce talent. Faced with this challenge, OpenAI had two options if it wished to meet its year-end objective: either obtain licences to use such trade secrets or reach other forms of agreement with manufacturers, or alternatively acquire the necessary knowledge through other means. It is possible that the latter route was chosen.
This explains why Apple’s complaint focuses on information assets such as metal alloys, finishing techniques, physical product designs, manufacturing methodologies, organisational processes, key supplier information, supply chain management, miniature systems design, energy consumption management systems, methods for transforming designs into products suitable for large-scale manufacturing, specialised machinery and production techniques, systems for reducing communications interference, and a long list of other assets. This clearly illustrates the remarkable versatility of trade secret protection. Those interested can consult Apple’s statement of claim for the full details.
With this litigation underway, it appears unlikely that OpenAI will be able to achieve its objective of launching its device by the end of the year. Should Apple ultimately succeed, any damages award against OpenAI could be substantial and might significantly hinder the company’s efforts to establish a hardware design and manufacturing division in the short to medium term.
Whatever the outcome, the case demonstrates the central role that trade secrets now play in modern industry. They enable companies to control critical assets that underpin competitive advantage and, where necessary, prevent competitors from benefiting from those assets unless they operate within the bounds of the law.
