The intersection of artificial intelligence (AI) and intellectual property (IP) is evolving rapidly, but it is far from straightforward. As highlighted during a recent WIPR event, the challenges in protecting AI innovation extend far beyond understanding the technology itself. With insights from industry leaders Douglas Gordon (Leonardo UK) and Sam Williams (Siemens), this blog explores key considerations for businesses navigating this complex landscape.
AI is "not one thing"
Douglas Gordon pointed out that we often talk about AI as if it’s a single entity. In reality, AI comprises multiple components including training data, algorithms and outputs. Each of these components might be protectable through different forms of IP but determining the appropriate protection depends on what the AI is doing, the business goals and the jurisdiction.
This fragmented nature of AI demands a tailored approach. IP must act as an enabler, not a hurdle, for business strategy. Questions like where is the product sold?, where is it manufactured?, and what does the local IP framework look like? are crucial to crafting an effective strategy. It’s the same IP toolbox, but applied in a new, more complex game.
The black box problem
AI is often perceived as a “black box” – a system whose inputs and operations are opaque, even to its creators. Sam Williams emphasised that this opacity poses challenges in securing patent protection. When inventors cannot clearly explain how an AI operates or what steps it follows, there’s a risk of patent claims being rejected.
Compounding this complexity is the dual role of AI: as both a tool for inventing and as the subject of inventions. What happens if AI is used to generate an invention disclosure? Or if it invents independently? Determining the relevant jurisdiction for such innovations is like opening Pandora’s box.
Engineering culture and IP discipline
Douglas Gordon highlighted a cultural hurdle: engineers often undervalue IP protection. They prioritise solving problems over safeguarding solutions, freely sharing ideas in the spirit of collaboration. While openness can drive innovation, it also exposes businesses to risks, especially when discussions happen without non-disclosure agreements (NDAs).
The solution isn’t to stifle creativity but to build IP awareness into the engineering process. Gordon drew parallels with health and safety practices which have become integral to engineering workflows. Similarly, “IP little and often” can ensure that ideas are protected without disrupting innovation.
Training is vital here, as Sam Williams noted. Managers, engineers and legal teams need tailored guidance on recognising and mitigating IP risks. For example, have teams been educated on copyright basics? Are IP assignments in contracts clear?
International challenges and opportunities
Both Gordon and Williams stressed the importance of a global perspective. While the UK Copyright, Designs and Patents Act (CDPA) provides guidance on who owns computer-generated works, this approach doesn’t extend worldwide. Protecting and enforcing IP across borders, particularly in jurisdictions like China and the US, requires careful planning.
China, for instance, has emerged as a leader in IP enforcement, especially for copyright. However, registering IP there can be challenging and requires a significant mindset shift in the registration and enforcement of relevant. Businesses need robust trade secret protocols to prevent sensitive information from leaking, particularly in team settings where knowledge is widely shared.
A pragmatic and layered approach
Protecting AI innovation demands pragmatism. Gordon advised businesses to adopt an “AI hedging strategy” – a layered approach to protection. By creating a series of IP “fences” around different components of AI, businesses can ensure fallback options if one form of protection fails. This approach also makes collaboration more cost-effective when disputes arise.
Trade marks and design registrations can play a complementary role in this strategy, protecting visual and brand elements of AI systems. However, the key is to keep revisiting and refining your AI IP strategy. As Williams observed, an IP strategy must be a living document that evolves alongside the technology and business needs.
What does this mean?
For businesses working with AI, the key takeaway is flexibility. Protecting AI innovation is not about choosing between patents, trade secrets, or copyright – it’s about using all available tools in harmony.
Understand the components: Training data, algorithms, and outputs each have distinct IP considerations.
Build IP into workflows: Train engineers and managers to treat IP as a routine part of their process, much like health and safety.
Think globally: Tailor your strategy to account for international variations in IP law and enforcement.
Be pragmatic: Use a layered, hedging approach to protection, accepting that some elements may fail.
Keep revisiting your strategy: AI and IP frameworks are evolving. Regularly update your approach to reflect new developments.
To find out more about the issues raised in this blog, contact Rosie Burbidge.