How to Patent AI Technology

AI technology like other similar computer-implemented technology can benefit from strong patent protection to protect the investment made in developing valuable products and services.  However, to avoid getting into difficulties with the various exclusions to patentability set out in the patent laws, it is important to focus on those aspects of the technology that are more suitable for patent protection.

Different countries have slightly different approaches to what is acceptable as a patentable invention and what is to be excluded.  As a general rule, patents should be directed to the technical aspects of the system, e.g. hardware, data processing steps, rather than the user aspects, e.g. display interfaces, user functionality etc.

Focussing on the apparent user benefits of a system in which an AI invention is implemented, can often give the impression of a less technical invention than is actually the case.  In order to maximise the likelihood of getting a good quality patent granted, it is important to bear the following general points in mind when considering what to try to protect.

  • Identify the core technical features of the system rather than the user interfaces (unless the interface is itself a technical improvement). Whilst a good user interface can help to maximise the commercial value and utility of an AI system, this can often fall foul of one the exclusions from patentability, such as the presentation of information.
  • AI systems may include complex algorithms based on multiple inputs possibly derived using machine learning. These can be difficult to define simply and so it may be necessary to try to identify their operation in a more functional language or try to break their function into smaller more definable portions.
  • Where the real effort has been in producing the model used to control a system, then it may be practical to protect the methodology used to derive the model rather than its use in a target system. This could be useful in preventing a competitor using the same technique to derive the same results.

Inventions relating exclusively to mental acts are difficult to patent, meaning it is important where an AI system replicates human action or activity, to focus on the technical elements that allow the invention to be implemented.  For example in a system using cameras and microphones for determining when it is safe to cross a road, steps human-like “looking for approaching cars”, “listening for the sound of vehicles” should be avoided but instead more technical steps like the process of analysing camera or microphone data as a technical process are more likely to be successful in a patent application.

Posted on: 12th April 2019