“Most people overestimate what they can do in one year and underestimate what they can do in ten years.” Bill Gates

Moore’s Law states that the number of transistors on a microchip doubles approximately every two years. For decades this has been a reliable predictor of computing advancements forecasting an exponential increase in computing power and a corresponding decrease in relative cost.

Now, the growth of Artificial Intelligence (AI) is outpacing Moore’s Law because the improvements in AI are dependent not solely on chip hardware but also on innovations in software, data availability and collaborative research efforts. Furthermore, the development of specialised AI hardware is also evolving rapidly, again exceeding the traditional boundaries set by Moore’s Law in terms of computational efficiency.

In contrast, Intellectual Property (IP) law is slow to evolve and struggles to keep pace with technological advancements.

To be effective, IP strategy needs to be responsive to rapidly changing technological and commercial environments. However, because IP law underpins most IP strategies, the currency and relevance of an IP strategy for AI based businesses which is relying on outmoded laws becomes questionable.

Here are some ways to navigate the changing landscape of AI and IP strategy.

Role Play

While it is easy to be bedazzled by the magic of AI, AI is just mimicking existing roles – except faster and with more data.

AI, or its outputs, should be looked at from the perspective of the role that AI takes on. Think of AI like any other third party which interacts with a business, and recognise that AI has similar attributes.

For example, during the creative process, questions can arise about an employee’s claim to have merely taken “inspiration” from another’s work. In the same way, AI is equally susceptible, if not more so, to accusations of sourcing its inspiration from materials created or owned by others.

In a similar vein, an employee can produce research that evidences poor fact-checking, while AI generated reports can be even worse at presenting inaccurate information.

Therefore, it is prudent to fact-check and independently ascertain sources when using information generated by AI. And indeed, greater care may be needed checking AI generated work than human generated work (depending on the human of course).

An IP strategy that recommends implementing internal systems that address these issues across human and AI generated work can mitigate embarrassing situations and potential litigation.

Authorship/Ownership/Inventorship

Currently, patent case law requires that an inventor must be human. Nevertheless, generative AI may be part of a creative process providing content and suggesting technical solutions. Questions however arise from this as to who the owner is or who can be attributed as the author/inventor. Is it the developer of the AI, the user who provided input, or the AI itself?

If the AI is considered as part of a collaborative process (like a group of researchers working together), then part of the answer to this question depends upon the respective contributions of each party. Regardless of how copyright or other IP laws evolve on this issue, applying best practice by recording contributions will provide information that can be used later to resolve any thorny attribution issues which might arise.

Market Approaches and Barriers to Entry

Traditionally, being first to market without any formal IP protection is not a strategy that works well – unless the market is niche and can be quickly satisfied. Often the unprotected first mover with a great idea is quickly overtaken by better resourced competitors who copy and improve upon that idea. Innovating faster and faster to stay ahead of the competition is not a sustainable nor a comfortable practice.

Less traditionally, but more commonly now, an AI supported business may depend upon gaining a critical mass of data for its learning models. Here, a first to market approach can work because at some stage the data accumulation and subsequent learning snowballs, with the first mover staying in front of the pack.

However, AI supported businesses are still vulnerable at the early market entry stage, when it is critical they gain traction quickly to take advantage of the snowball effect. A considered IP strategy can help here. Barriers to entry can be created – even just temporarily to give competitors “Cause to Pause” – providing the head-start needed.

For example, an unpublished patent application may prompt competitor hesitation until the application is published (usually 18 months after being filed) and the competitor can assess the likely scope of any patent that may be granted.

Other barriers to entry that may impact IP strategy include regulatory frameworks. For instance, data protection regulations like the GDPR, affect how data can be used for AI training. Emerging AI-specific regulations may introduce new compliance requirements, influencing the development and commercialisation of AI technologies. A first mover gaining regulatory approval can have a significant competitive advantage.

Competitor hesitation around barriers to entry may be all that is needed for a business to gain market domination. It may not even be necessary to continue with formal IP protection such as patents. Having an IP strategy that is informed by and incorporates the complete AI landscape, market, and regulatory and IP issues, helps immeasurably in enabling traction in the market.

Patentability

To be patentable, an invention must be novel, inventive, and useful. This criteria also applies to AI inventions – with the added complication that some jurisdictions such as Europe and China, have a more rigorous approach to patentable subject matter than other countries like USA and Australia.

When using publicly available machine learning models, it can be difficult to determine what AI supported inventions could be patentable, given that there is so much in the public domain.

Helpfully, the USPTO has provided guidelines around the structure of AI patent claims that could be accepted for grant – essentially requiring that a process be clearly outlined – as given in the example below.

Patent attorneys who work with innovators and train them to pull out inventive aspects conceptualised in the manner of the USPTO guidelines, are invaluable for maximising the protectability of AI inventions.

As the protectability of software and AI is rapidly changing, always check with an IP professional before deciding on a protection strategy in this field.

Security, Trade Secrets and Data

IP strategies usually have a multi-pronged approach. This is especially appropriate with the uncertainty around whether formal protection mechanisms (such as patents) can apply to AI and its outputs.

Data is the lifeblood of AI. Its ownership and protection are critical, and this can raise issues such as the legality of using data for training AI models, data privacy concerns, and the rights to data-derived insights. Establishing clear data ownership and usage policies is essential for protecting IP.

AI supported companies often rely on trade secrets to protect their proprietary algorithms and data. Trade secrets can provide indefinite competitive advantage as long as the information remains secret. However, maintaining secrecy requires robust internal policies and security measures.

An IP strategy that combines formal (and eventually published) protection with good internal systems will obviously benefit most AI based businesses.

Final thoughts

Understanding the curveballs thrown by AI and how they can affect the development of IP strategies is essential for fostering innovation, securing competitive advantages, and navigating a complex legal landscape. By addressing the unique challenges posed by AI, and front-footing these in an IP strategy, businesses can better control their innovations and capitalise on the opportunities presented by this transformative technology.