Last month, Elon Musk officially launched the generative Artificial Intelligence (AI) startup, xAI. Tasked with a goal to “understand the true nature of the universe”, the team at xAI consists of a number of engineers previously associated with leading actors in the generative AI space, including DeepMind and OpenAI.
This high profile launch of yet another AI startup is further evidence of the ongoing interest in the potential of AI (both technologically and commercially) and the skyrocketing investment the sector has seen. Unsurprisingly in this context, companies are seeking to protect the Intellectual Property (IP) that is rapidly being developed by their AI engineers.
Over the past decade, there has been vast growth in the number of patent applications filed for inventions relating to AI despite there currently being a number of challenges associated with the patentability of such subject matter in some jurisdictions. For example, in Europe, applicants may face objections that their AI inventions are excluded from patentability and/or devoid of technical character. Such objections are not irrefutable (as evidenced by the number of patent applications being granted for AI subject matter), but they can complicate and lengthen the patent prosecution process. Nevertheless, the filing statistics in this area are a strong indication of the commercial value of patents for companies innovating in AI.
Of course, patenting an invention requires its workings to be disclosed to the public. Thus, as seen with OpenAI and their decision to withhold the technical details of their more recent Large Language Models, some companies are becoming more heavily reliant on trade secrets and the perceived extreme difficulty of reverse engineering AI models. However, as discussed in our article reviewing the detectability of AI patent infringement, it may not be safe to assume that the complexity of AI models means that they cannot be reverse engineered. Deciding to forego the public disclosure of AI models (which is a part of the patenting process) in order to maintain a competitive advantage is not a risk-free strategy.
Even when discounting the risk of reverse engineering, it is important to consider that competitors may arrive at the same AI model via independent creation. In addition, the future regulation of AI – a much debated topic in 2023 – will likely limit the ability of actors in this field to completely prevent public scrutiny of the AI models they employ. In particular, there appears to be some consensus among law makers that AI models and their algorithms should be made transparent for regulatory purposes when it is in the public’s interest or relevant for public safety.
Of course, we could argue that this requirement for increased transparency imposes an obligation for IP policy makers to move towards a more permissive approach to the patent eligibility of AI inventions. For example, if AI innovators are to be expected to publicly disclose the results of their costly R&D , considering AI subject matter as inherently eligible for patent protection would help to protect the commercial value of that investment.
Crucially, with the strong growth in both AI investment and AI patenting activity, regulation on the horizon and AI rarely out of the news, it has never been more important for innovators to carefully consider their strategies for protecting and monetizing AI inventions within both the current and future commercial landscapes.