While looking through some social media feeds last week, I came across Seedlegals. Based in the UK, it uses chat bots and an app to interface with companies seeking corporate legal advice. The system constructs terms sheets and deal documents, as well as provides answers to some questions concerning corporate formation. This is not a UK-only phenomenon. Shoobx, a web-based service, provides incorporation documents for Delaware company formation, as well as equity management and human resource tools.
So, I got to thinking, how would this work for intellectual property advice. What if I made an IP chat bot?
And on it would go, discussing the new invention, looking at prior art and then the app would spit out a new patent application or maybe even file it on the spot. Really? No, probably not.
At least at this basic level, such a system seems too generic and templated. Now there are some legal tasks that lend themselves to cut and paste-type forms. Some agreements are very similar and used often, differing only on specifics like party names, duration and price. But other tasks require a more tailored approach.
A patent application is one of these more tailored undertakings. Sure, there can be some shared core elements like definitions. However, for a company, particularly a start-up, a patent application isn’t a stand-alone activity. The application is part of a higher-level strategy, integrated with business goals and timelines, as well as gaining traction in the competitive landscape.
There are other questions that arise when thinking about how a chat bot or AI-based app would provide legal advice. For example, would the bot have attorney-client privilege? It’s not an attorney. Would it be treated as an agent of an attorney? Otherwise, all conversations could be discoverable, including statements an applicant made to the bot about prior art and the scope of the invention. Would the chat bot or app have malpractice insurance? Perhaps this would be carried by the bot’s “owner.” However, how might this play out, particularly if the app is autonomous in its client interactions and the AI is designed and trained by a third party?
Kidding aside, I am not trying to discount machine-learning and artificial intelligence-based tools for IP. I do think these tools can have a role in augmenting the capabilities of an attorney.
To date, I’ve seen companies advertising “AI-based” patent tools. There isn’t enough in the promotions to gather whether this is truly AI with the algorithm training and learning on a known data set then capable of making predictions on new data, or whether the systems are just high-powered search engines combing through high volumes of data.
For instance, I have seen offering for compiling all worldwide patents and applications using related key words. These searches can be mapped into a landscape or turned into bar graphs showing trends for country of filing, patent term, assignee, inventors and the like. While useful for gathering information and sorting it, these implementations do not provide much in the way of analysis. The volume of data they display can often be too high and the relevance and utility for the bulk of the data is questionable.
With AI, the systems should be capable of providing a more robust and “thoughtful” deliverable. For example, construct a patentability predictor for the USPTO. The system could ingest all of the patent office actions, rejections and responses to applicant arguments in a related field. The system could integrate this with changes in the current case law from the relevant courts (Supremes, Federal Circuit and district courts). It would need to take into account the timeframe as well, since patent law and its practical implementations evolve over time. Could the system then provide a rating for likelihood of patentability for similar technologies examined by the same art unit or the same examiner? This could be used by the attorney as a starting point for generating an effective claim strategy and response to the examiner’s rejections.
Another useful tool for the future could assist an attorney in addressing freedom to operate. Based on training data, the algorithm when fed a product, a patent at issue and its file history, could identify potential claim construction positions and prosecution estoppel issues and provide an initial infringement assessment that could be further refined and deepened by the attorney. On the validity side, the tool could find prior art and in addition, construct an initial claim chart matching the prior art elements from single and combinations of references to the claims at issue.
I can’t say I’ve seen anything this sophisticated so far. I’ve run across services that claim to automate patent drafting with some type of AI/machine learning tools. My initial take is that these services remain very simplistic, currently capable of looking at wording issues and the like (such as a word in a claim that doesn’t appear in the patent’s specification or a missing antecedent basis in a claim). Again, I don’t dismiss the usefulness of the tools, but they seem to be aimed at present to the mass production of patents.
In my view, if we are going to promote AI-based tools in the industry, let’s raise the bar and aim high. It’s not volume we are going for, and not even cost alone, but quality and strategic value. The patent applications you file are intended to protect your products and technology platforms. The idea isn’t to just get any patent to issue, but to get one that will be valuable to your business strategy, with sufficient breadth and validity to stave off competitors.