Following a recent announcement by the EPO, a preview of the next version of the Guidelines for Examination is now online. One of the changes highlighted was a section on the patentability of AI and Machine Learning inventions.
While the new section is detailed and provides some helpful examples, the content is very much as expected: real-world technical applications of AI and ML are patentable, as are inventions relating to inventions specifically adapted computer hardware. However, it would appear that any fundamental advances in AI and ML that are not specific to any application or hardware remain unpatentable.
Why is this the case? Why is ML, fundamentally techniques for making a machine learn (adapt itself) to represent and extract patterns from data not considered technology when applied to data as such? This seems strange given that in other fields, such as cryptography, the application of abstract algorithms to generic data is patentable. Granted in one case the data is secured and in the other case a machine is enabled to do useful things with the data but why does this make one case technology and the other one not? It seems inconsistent to discriminate against a whole field of technology in this way.
Amongst much hand-wringing that patenting AI and ML inventions as such would require a change in legislation, it appears to this writer that a recognition of what AI and ML actually is all that is needed.
For more information or advice, please contact Alexander Korenberg or your usual Kilburn & Strode advisor.