US Patent Office Changes Rules for AI-Assisted Inventions

The US Patent Office (USPTO) surprisingly withdrew its previous guidelines on the patentability of AI-assisted inventions on Wednesday and replaced them with a new set of rules. The central message from the agency under Director John Squires is: Generative Artificial Intelligence systems like ChatGPT, Gemini, or Claude are to be considered "analogous to laboratory equipment, computer software, research databases, or any other tool" in the invention process.

This clarifies that inventions made with the help of AI are fundamentally patentable – as long as a human is in control. With the new document, the USPTO is completely repealing the guidelines that were only introduced in February 2024 under the Biden administration. The old guideline suggested evaluating AI contributions based on the so-called "Pannu factors": a legal standard, which is actually intended to distinguish the contribution of multiple human co-inventors.

Squires now dismisses this approach as illogical. Since AI systems are not natural persons, they can legally never be considered "Joint Inventors." Consequently, it is unnecessary to apply standards for computer programs. "There is no separate or modified standard for AI-assisted inventions," the notice states. The same law applies as for any other invention.

The Human Must Have the Idea

The linchpin for a successful patent application remains the intellectual creation in the inventor's mind. The USPTO clarifies: Artificial intelligence may provide services and generate ideas. However, it can never be named as an inventor or co-inventor in the patent application. The decisive factor is whether a natural person has formed a "definite and permanent idea of the complete invention" in their mind.

Similarly, the European Patent Office has argued for years. Software and computer systems "as such" are not protectable under the European Patent Convention. Patents on AI and algorithms are therefore generally highly controversial in Europe.

According to the USPTO, an inventor must be able to describe their innovation so precisely that it can be put into practice without excessive experimentation. If an AI merely spits out a vague idea and the human does not intellectually penetrate and define it as a concrete solution to a problem, it is not sufficient for a patent.

Consequences for International Applications

The guidelines apply to all types of patents, including design and plant patents. The Office warns of problems with international priority claims: anyone who files a patent application abroad naming an AI as the sole inventor cannot claim priority in the US. In the United States, only applications naming at least one natural person as inventor will continue to be accepted.

The new regulations are intended to end the legal uncertainty that has prevailed when researchers and developers used AI as an intensive development partner in the innovation process. They will now be treated legally as if they had used an advanced microscope or complex simulation software.

(nen)