If you came just for the checklist: Download Here »
The other day I came across this post about a claim drafting seminar my IP friend Ronak was giving. Interestingly, most attendees were in-house patent attorneys who wanted to be able to better check the quality of the drafts delivered.
This made me think:
How can an in-house person check the quality of a software patent claim draft delivered by their outside counsel?
Here are five quick checks I’d do:
Check no. 1: Is claim 1 a method claim?
Typically, the inventive gist in a software patent lies in its functionality, i.e., in its runtime behavior. The natural claim category is thus a method claim that sets out the steps performed by the software.
I’ve seen some European patent attorneys saying that a claim “looks more technical” if it’s a system or device claim with an input interface, a processing module, etc.
But in my experience such structural elements only tend to add fluff to the claim which does not actually help for patentability.
Even worse, they might give an infringer room to argue that their software has a different structure.
Therefore, I’ll go with a method claim as claim 1 in maybe 95 % of my patent applications.
Check no. 2: Does the claim set include at least three independent claims – method, device, computer program?
See the EPO Guidelines F-IV, 3.9.1 for recommended claim wording in these three basic categories.
Each claim category comes with its own set of infringing activities:
The method and device claims are oftentimes directly infringed only by the customer, i.e., the user of the device.
But the computer program claim is typically directly infringed by the competitor, i.e., the provider of the software.
Check no. 3: Do the independent claims capture the “empty” software product?
For example, a claim with a “database storing a plurality of entries” might only be directly infringed by the customer who populates the database during use.
But the competitor might only provide the empty database. A better claim wording might thus be “database configured to store a plurality of entries”.
Notice the subtle change? It can make all the difference.
Check no. 4: For distributed systems: Does the claim set capture each entity individually?
Think about a data producer and a consumer.
A client and a server.
A machine and a digital twin.
Make sure the claim set includes individual independent claims for each relevant entity.
The competitor might only be responsible for one participating entity (e.g., the server), i.e., only the entity claims are directly infringed whereas a “whole system” claim would at best be indirectly infringed.
Of course, each independent claim must be patentable on its own, which might not always be the case. See the EPO Guidelines F-IV, 3.9.3 for recommended claim wording.
Check no. 5: For AI inventions: Are there separate independent claims for the training and for the use of the machine-learning model?
Oftentimes, the machine-learning model is trained by one entity and then provided to another entity which uses the trained model at runtime.
A claim mixing the training and use might at best be indirectly infringed by each of these entities, so make sure the claim set includes two independent claims in this example.
Of course, each independent claim must be patentable on its own, which might not always be the case.
Need a handy checklist?
I’ve compiled these quick checks into a handy checklist as a free download. It also includes additional test questions to ask when you compare the claim set with the invention description.
Hope it helps!