- within Compliance topic(s)
For years, the debate surrounding artificial intelligence and patents has focused on a relatively simple question: Can AI be an inventor?
At least in the United States, the answer is currently no. Inventors must be natural persons.
Problem solved. Or perhaps not.
While lawyers, courts, and the USPTO have spent the past several years debating whether AI can be listed on a patent application, a more interesting question has quietly emerged.
What happens when AI does not invent the technology, but instead invents the innovation strategy?
Patent law revolves around the concept of conception. The Federal Circuit has repeatedly described conception as the touchstone of inventorship. Put simply, an inventor must possess a definite and permanent idea of the complete invention. The inventor must know what the invention is before reducing it to practice.
The question becomes more complicated as AI systems evolve from research assistants into innovation engines.
Researchers are developing platforms that combine methodologies such as TRIZ (Theory of Inventive Problem Solving), graph databases, and multi-agent AI frameworks. Rather than simply answering questions, these systems analyze existing technologies, identify technical contradictions, generate potential solutions, predict future development paths, and even recommend patent strategies.
Imagine an engineer working on battery technology. The engineer asks an AI system how to improve charging efficiency without increasing heat generation. The AI analyzes thousands of patents, technical papers, and product specifications. The AI identifies an overlooked technical relationship, proposes a novel architecture, predicts performance improvements, recommends claim language, and suggests a continuation strategy to protect adjacent implementations.
The engineer reviews the proposal, recognizes its value, and instructs the engineering team to build it.
Who conceived of the invention?
Under current law, the answer may still be the engineer. After all, the engineer identified the problem, evaluated the solution, and directed further development.
But the facts begin to look different from the traditional conception story that patent law has relied upon for decades.
Historically, the inventor had the idea and used tools to implement the idea. Increasingly, the tool may be generating the idea while the human decides whether the idea is worth pursuing.
The USPTO’s Answer: AI Is a Tool
In November 2025, the USPTO revised its guidance on AI-assisted inventions and largely simplified the analysis. The Office rescinded its earlier guidance and explained that there is no special inventorship standard for AI-assisted inventions. The same legal principles apply regardless of whether an inventor used a laboratory instrument, a computer program, a research database, or an AI system.
The USPTO’s position is straightforward: AI systems are tools.
According to the guidance, AI is analogous to software, laboratory equipment, or other instruments that assist human inventors. AI may provide information, suggestions, and even ideas, but AI itself cannot be an inventor. The relevant inquiry remains whether a natural person conceived the claimed invention.
For today’s AI systems, that approach makes a great deal of sense.
The challenge is that tomorrow’s AI systems may look very different from today’s.
The Real Question
The recent USPTO guidance treats AI as a tool, and for today’s systems that characterization is entirely reasonable.
But the guidance also highlights a deeper issue.
Patent law was built on the assumption that conception occurs in the mind of a human inventor. The doctrine works well when humans generate ideas and machines help execute them. The doctrine becomes more difficult to apply when machines increasingly generate ideas and humans decide which ones deserve attention.
The harder question is whether future AI innovation platforms will remain merely tools, or whether they will become something patent law has never encountered before: systems capable of generating the very conception that inventorship doctrine was designed to identify.
The USPTO has answered today’s question by confirming that AI is a tool, not an inventor.
Whether that answer remains sufficient ten years from now is far less certain.
And if that day arrives, the most important person in the inventive process may not be the engineer, the patent attorney, or even the inventor.
It may be the person who wrote the prompt.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
[View Source]