Patents for AI & Machine Learning
AI inventions often sit between software, math and hardware. Framing them well matters.
What tends to matter
Software and AI claims are examined closely under US eligibility law, so how the invention is described affects its prospects. We focus on the technical problem solved and the specific technical improvement: how data is represented, how training or inference is made faster, more accurate, more robust or more efficient, and how the system interacts with real hardware or processes.
Typical subject matter
- Model architectures, training methods and data-handling pipelines
- On-device and edge inference, compression and acceleration
- Computer vision, sensing and signal processing, including medical imaging
- Analytics and business-intelligence systems built on learned models
- Systems that combine models with control, robotics or devices
Related domains in Metayage’s patent showcase
Metayage, the company behind CanPatentIt, organises its patent showcase by technology domain. The domains closest to this area are:
Common questions
Can an AI model or algorithm be patented?
It depends on how the invention is framed and what technical improvement it provides. We assess this case by case; no outcome can be promised.
Information on this page is general and not legal advice. Outcomes cannot be guaranteed.