For a recent article published in Law360, we looked at data from the US Patent and Trademark Office (USPTO) covering both AI and non-AI inventions from 2015 to 2020. Our findings show a stark difference: AI-related inventions were rejected for not being eligible subject matter 45% of the time, compared to just 10% for non-AI inventions.
To qualify for patent protection under US law, inventions must demonstrate “subject matter eligibility,” meaning they must have a concrete and practical application. AI inventions have had higher rejection rates in part due to the abstract nature of AI technology, according to USPTO interpretations of current patent guidelines.
We also examined how different types of AI technologies fared. The USPTO categorizes AI into eight components. AI inventions categorized as "evolutionary computation" faced the toughest odds, with 58% of applications rejected. Those categorized as "vision" technologies had the easiest time, with only 33% of applications rejected.
[View source.]