ChatGPT. DALL·E 2. These are two names, amongst many others, which are at the tips of everyone’s tongues these last few months, and rightly so, with the immense potential shown by the numerous disruptive developments in the realm of Artificial Intelligence. Such AI has shown the masses an entirely new way of how to interact with the digital world, and how to extract results. 

While the existence of Artificial Intelligence is hardly new, the developments in the last 12-18 months, and especially its widespread proliferation and use by way of software such as ChatGPT has caused a paradigm shift in how we interact with the internet. Various industries have wholly positively embraced and adopted such AI, while many have been quick to reject the utility of AI in their industry. The Legal Industry, which many would perhaps stereotype as conservative, has been quick to realize the vast potential of such AI, and as such, has not only positively embraced the technological developments, but taken great strides in trying to implement such AI in their day-to-day functionalities. 

Applications of AI in patent law

One of the many highly debatable applications of such AI to the legal industry, is with respect to drafting of patent applications. Quality notwithstanding, AI like ChatGPT can be used in the following manner when it comes to patents – 

  1. Generating patent claims; 
  2. Preparing iteration of claims; 
  3. Preparing description of inventions;
  4. Provide patent specifications; 
  5. Compare patent claims; 
  6. Patent Searches;
  7. As a dictionary, while drafting; 
  8. the current state of the art; etc.

All the above tasks, if mastered by AI, can lead to numerous benefits for Patent Agents/ Attorneys and Inventors alike, including but not limited to: 

  1. Reduction in patent prosecution fees;
  2. Reduction in time taken for drafting; 
  3. Templatization of drafting, at least to a certain level; 
  4. Ease of drafting for beginners; 
  5. Inventors can further fine-tune their invention by seeing how the basic claims work out by using AI, without engaging a patent agent/ attorney. 

Thus, on the basis of the current level of AI technology, the possible potential of AI in the realm of patent prosecution is huge. 

Usage of AI in patent prosecution – realistic concerns

While it is understandable to be excited about the many possible applications of AI in patents, there are quite a few concerns about the same as well. For instance: 

  1. Privacy Concerns: Perhaps the most severe risk posed by AI such as ChatGPT in this context, is privacy. Such AI are based on continuous learning. In layman terms, such AI constants learns on the basis of the human input provided by users. This is in direct contradiction to the core tenet of patents, which is novelty. For instance, there is a very threat that inputting details of an invention prior to patenting might lead to crucial details being assimilated by the AI, which can then turn up when queries related to similar inventions are put forward by third parties in the AI, thereby potentially risking the very existence of the novelty. 
     
  2. Accuracy of Claims Drafted: There was a recent case in the USA, wherein a lawyer utilized ChatGPT to quote precedents in a case. However, the so called precedents turned out to be completely made-up![1] Thus, the accuracy and reliability of AI can be questionable, and as such, can completely jeopardize legal proceedings. This is exacerbated by the fact that the claims which are drafted using AI may prima facie sound plausible, which blurs the line between true and false. 
     
  3. Ensuring Novelty In Claims:[2] AI may come up with technically and scientifically correct claims, but may lose out on one of the most important tenets of patent law – which is the inventive step. The job of a patent drafter is to ensure that the invention is non-obvious and has an inventive step, over prior existing inventions. Thus, the usefulness of AI to patent practitioners would also be dependent on whether the claims being churned out by the AI include aspects which are non-obvious and count as inventive step. 
     
  4. Sources:[3] For AI to be truly helpful in drafting claims, the database should have access to a large number of patent databases. Afterall, AI gets trained by using the general data and information available over the internet. However, in order to on a specific patent and draft, any skilled human practitioner would invest a lot of time and energy in the research, including research in various patent databases. However, if one were to utilize AI, the AI would be accurate and actually help in a constructive manner if the AI has access to a large number of patent databases, and is thus abreast with current events or specific patents that could be relevant to an application.
     
  5. Patent Strategy: Even if the AI comes up with technically astute and correct patent application, the AI lacks subjectivity. For instance, in many cases, a skilled patent practitioner may make small changes to the draft, even at the cost of narrowing down the claims, for important reasons such as – 
    -- Avoiding infringement;
    -- Maximizing the potential for monetization; etc
     
  6. Possible decrease in efficiency? In view of the above teething problems, a consequence of usage of AI in patent drafting is that while the AI may churn out plausible sounding output, if the output is fundamentally incorrect, the same may take a patent agent/ attorney quite some time to basically reverse-work on the draft and make sure that the final patent document would be a proper one. While subjective, it is possible that many patent agents/ attorneys might be able to more quickly draft from scratch, as compared to using AI tools. 

The takeaway

It is undeniable that AI is here to stay, and that it is going to become an important part of the legal industry, especially in the field of patents. However, while the potential is indisputably there, and in fact there are tangible current usages of AI in patent drafting, one must be cautious while using the same, and keep in mind the above realistic concerns, especially those pertaining to data security and maintaining the integrity of novelty of an invention. Especially while using AI which rely on Large Language Model (LLM), it is prudent to ensure that the AI is not being inputted with data which, if accessed by third parties, would basically amount to loss of novelty. For instance, while using LLM based AI, the user might wish to utilize it to only do the grunt-work and not use it for the core-drafting, which requires the inputting of confidential data. Further, the practitioner should also check what all patent databases are covered by the AI, and accordingly manually refer to important databases, which aren’t covered by the AI, to ensure the integrity of the patent draft. Thus, given the current state of the technology, patent practitioners must treat AI as a limited tool rather than a crutch. 

That said, like any other disruptive technology having such massive potential and applications, once such problems are identified, it is only a matter of time before the concerns are addressed. 

[1] https://www.bbc.com/news/world-us-canada-65735769 
[2] https://www.jdsupra.com/legalnews/can-chatgpt-draft-patent-applications-7924300/ 
[3] https://www.law360.com/legalindustry/articles/1586838/the-perils-and-promise-of-chatgpt-as-a-patent-drafting-aid 
 

BECOME A MEMBER OF FICPI

FICPI is the only international NGO whose membership consists exclusively of IP attorneys in private practice. As a result, the FICPI community is driven by a shared interest to promote solutions and advocacy for private IP practitioners.
BECOME A MEMBER OF FICPI
MEMBER LOGIN