Today, I would like to provide you with some guidance on how to avoid three typical mistakes that I frequently encounter in patent applications relating to AI-based inventions.
These mistakes are: 1) Divided infringement, 2) insufficiency of disclosure, and 3) lack of plausibility
So, how can these mistakes be avoided? In practice, an AI model may be trained by a first entity, whereas the trained model may then be used by a second entity. Hence, a claim that is directed at both training and using a model may be neither infringed by the first nor by the second entity. To avoid such case of divided infringement, training a model and using the trained model should be claimed separately. This will significantly simplify asserting your patent towards potential infringers.
To meet the requirement of disclosing your invention in a manner sufficiently clear and complete for it to be carried out by the skilled person, it might be helpful to supplement your application with some sample training data, details on the model used, and details on the training method. This may enable the skilled person to create their own set of training data that can be used for training a model according to your invention.
Finally, to make it plausible that your invention actually produces the alleged technical effect, I suggest incorporating some test or measurement results into the application that show how well your AI-based invention operates.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.