### The Ghost in the Machine: USPTO Rules AI Can Assist in Invention, But Humans Must Take the Credit

For years, a question has loomed over the world of innovation: if an artificial intelligence system creates something new and useful, who gets to be the inventor? The U.S. Patent and Trademark Office (USPTO) has finally provided a clear, if not final, answer. In new guidance, the agency has declared that while AI can be a crucial tool in the inventive process, it cannot be named as an inventor on a U.S. patent. The credit, and the legal inventorship, must go to a human being.

This landmark guidance addresses a growing gray area in intellectual property law, accelerated by the rapid advancement of generative AI in fields from drug discovery to engineering. The core of the USPTO’s position rests on the current interpretation of U.S. patent law, which states an “inventor” must be a “natural person.” An AI, no matter how sophisticated, does not qualify.

So, where does that leave the thousands of researchers and companies integrating AI into their R&D workflows? The guidance clarifies that an invention created with AI assistance is not automatically unpatentable. The key is determining whether a human has made a “significant contribution” to the invention.

According to the USPTO, merely owning or operating an AI system isn’t enough to claim inventorship. A person must have contributed significantly to the “conception” of the invention. This can be demonstrated in several ways:

1. **Constructing the Prompt:** A human who designs a specific prompt that elicits a solution from an AI system could be considered an inventor. This goes beyond simply asking a general question; it involves a creative and targeted query that steers the AI toward the invention.
2. **Designing the Experiment:** The human who designs an experiment or model that the AI uses to generate its output can be credited with a significant contribution.
3. **Recognizing and Appreciating the Output:** Perhaps most crucially, a human must be the one to recognize the significance of the AI’s output and understand its potential as an invention. An AI might produce thousands of results, but it’s the human expert who identifies the breakthrough.

This guidance is a direct response to high-profile legal challenges, most notably the case of Stephen Thaler and his AI system, DABUS (“Device for the Autonomous Bootstrapping of Unified Sentience”). Thaler filed patent applications worldwide, listing DABUS as the sole inventor. While a few jurisdictions, like South Africa, granted the patent, major bodies including the USPTO, the European Patent Office, and UK courts rejected the application on the grounds that an inventor must be human. The U.S. Supreme Court ultimately declined to hear the case, letting the lower court’s decision stand.

For innovators, this new clarity is a double-edged sword. On one hand, it provides a viable path to patenting AI-assisted inventions, protecting their investments in AI-driven R&D. Companies now have a clearer playbook: document human involvement meticulously. Keep detailed records of who designed the prompts, how the AI was trained for a specific purpose, and who made the critical decision to pursue the AI-generated result.

On the other hand, it raises questions about the future. What happens when an AI becomes so advanced that its “contribution” vastly outweighs any human input? If an AI, with minimal prompting, conceives and validates a new chemical compound entirely on its own, the current rules suggest it might not be patentable at all if no human can claim a “significant contribution.”

The USPTO’s guidance is a pragmatic solution for the technology of today, grounding the law in the familiar concept of human ingenuity. However, it also sets the stage for a much larger conversation. As AI evolves from a sophisticated tool into a potential creative partner, lawmakers may eventually have to revisit the very definition of “inventor.” For now, the ghost in the machine can help create the future, but a human will be signing for it.