Protecting AI diagnostics – two bites at the AIpple?Posted on
The use of AI has revolutionised the speed and accuracy of diagnosis for a host of conditions including breast cancer, prostate cancer, digestive diseases, skin cancer, lung cancer, neurodegenerative diseases, diabetic retinopathy, heart disease and stroke. That in turn can lead to earlier diagnosis and timely targeted intervention, ultimately improving survival rates.
So, AI diagnostic inventions are looking after us – but how can we protect them?
Typically, protection for an invention is via a patent or a trade secret. But which is best for AI diagnostic inventions?
The first consideration is whether the invention is eligible for patent protection. If that hurdle is not cleared, trade secret protection is the obvious route.
AI diagnostics can relate to several areas excluded from patent protection per se, for example mathematical methods, computer programs, or methods of diagnosis. However, the UK and European patent offices consider those exclusions to be avoided by an automated system processing physiological measurements to provide a medical diagnosis – such as the use of a neural network in heart monitoring apparatus to identify irregular heartbeats.
Patent or trade secret?
The role of a patent is to protect your commercial position by preventing third parties from using your invention. In return, however, the patent must disclose the invention, in sufficient detail for a skilled person in that field to put the invention into practice. That would prevent trade secret protection, which as the name suggests requires the invention to be kept secret.
The next consideration therefore is whether patent protection is worth the disclosure of your invention? This is a multi-faceted question that depends on the nature of the invention itself, such as how easy it is to reverse engineer, or what your commercial goals are.
However, for some inventions the choice between patent protection and trade secret protection is not always clear.
The good news for such inventions is that the choice between patent protection and trade secret protection need not be made from the outset. It is sometimes possible to explore one form of protection without jeopardising the other.
This is because patent applications are not made public until approximately 18 months after the filing date, allowing inventors a period during which they can decide on the best course of action. It is also important to mention that depending on the jurisdiction, it is possible to obtain a preliminary opinion on patentability from some patent offices prior to publication.
Putting protection into practice
One route could be to file a patent application first. If initial feedback from a patent office (for example search results or a preliminary opinion) indicates broad and commercially valuable patent protection for the invention looks possible, the application can be allowed to publish and prosecuted through to grant.
If not, the application could be withdrawn prior to publication. If the application is not published, trade secret protection may still be relied upon.
Alternatively, trade secret protection could be employed initially.
If the AI algorithm is effectively a black box, it is possible that commercialising it will not be considered an enabling disclosure that could jeopardise patentability. This route could keep open the opportunity to pursue patent protection, if necessary, at a later date. If the trade secret protection is compromised for any reason, there are also provisions in the UK and Europe allowing patent applications to be filed after breaches of confidence.
There are also some general – and by no means exhaustive – considerations for determining whether an invention may be best protected via a patent or a trade secret that inventors should weigh up.
Most importantly AI algorithms are typically difficult to reverse engineer, so it may be preferable to avoid disclosing specific detail regarding the algorithm itself to any competitors in a patent application. However, trade secrets do not protect against independent third-party developments.
Another aspect to consider is that broader patent protection may be available for inventions where the innovation lies outside the AI algorithm itself, for example if the invention lies in pre-processing steps or a combination of input parameters that improves the output of the AI. In those cases, patent protection may not be limited to a specific AI algorithm, which may make the patent harder for third parties to design around.
If your invention instead lies within the AI algorithm itself (for example a particular neural network structure that is well-adapted for a particular diagnostic challenge), the patent may be easier for third parties to design around. Consideration should also be taken on the likelihood of the patented technology becoming obsolete due to the speed of development in AI algorithms.
If you would like to discuss patent or trade secret protection for your AI diagnostic invention, please contact the author or your usual Barker Brettell attorney.