USPTO Eases Patent Eligibility for AI and High-Tech Inventions, Opening New Opportunities for Inventors

The U.S. Patent and Trademark Office (USPTO) has signaled a more welcoming approach to patenting artificial intelligence, software, and other high-tech inventions, reversing a trend of strict rejections that had discouraged many innovators. Over the past several months, the agency has issued new guidance and a precedential decision that clarify when such inventions qualify for patent protection, offering renewed hope for inventors in fields like electronics, automotive technology, and medical devices.

Under USPTO Director John A. Squires, who took office in September 2025, the agency issued the Appeals Review Panel decision Ex parte Desjardins in late September 2025. That decision vacated a rejection of machine-learning claims, holding that improvements to the functioning of a machine-learning model can constitute a patent-eligible technological improvement rather than an unpatentable abstract idea. As summarized by intellectual property analysts, the USPTO followed with examiner guidance directing that claims should not be dismissed at too high a level of generality and that artificial-intelligence systems should be treated as tools used within an invention.

Director Squires has described the USPTO as open to transformative technologies, including artificial intelligence, quantum computing, cryptocurrency, and medical diagnostics. However, the shift is not a blank check. The new guidance binds patent examiners, not the courts, and the Federal Circuit continues to apply the same eligibility framework. This means a patent that clears examination can still be challenged in litigation if its claims merely apply generic computing to a familiar idea.

For inventors, the recalibration means the USPTO is now less likely to reject AI, software, and computer-related claims under Section 101 as "abstract," particularly when the invention improves how a computer or technical system actually works. Recent guidance narrows the "mental process" rationale often used to reject machine-learning claims, recognizing that models processing large volumes of data operate beyond human cognition. Inventors who were previously told an idea was "too abstract" to patent may want to revisit that decision under the current framework.

Novelty, non-obviousness, and disclosure requirements remain unchanged, so a strong application still depends on careful drafting. Because the courts have not changed course, a specification should clearly describe the specific technical problem being solved and how the invention improves the underlying technology, so a granted patent can hold up if later challenged. The change reaches beyond pure software, touching connected devices, electronics, automotive and transportation systems, and diagnostic technologies that depend on data and machine learning.

"For a decade, a lot of good engineering never made it into a patent application because everyone assumed it would be rejected as too abstract," said J. Baron Lesperance, founder of The Patent Baron, PLLC. "That calculus has changed, and inventors in electronics, automotive tech, and medical devices should take a fresh look." Based in Michigan, The Patent Baron, PLLC brings an engineer's eye to intellectual property, with founder and patent attorney J. Baron Lesperance pairing legal training with graduate degrees in electrical and computer engineering and years in industry. Inventors curious whether the USPTO's evolving stance opens a path for an idea they once set aside can schedule a consultation with the firm.