AI Is Patentable Again, For Now

A fast sequence of USPTO actions has reopened the door for AI and software claims. The reprieve is real but it is not permanent. Here is why patent professionals should not give up on pending AI applications this quarter.

For most of the last decade, filing an AI or software patent application felt like a wager. The Alice/Mayo framework turned Section 101 eligibility into a moving target, and examiners in the software art units built a habit of labeling almost any algorithmic advance an “abstract idea” or a “mental process.” Applicants learned to expect rejection first and argue later. That habit is now breaking and it is breaking quickly.

In roughly four months, the United States Patent and Trademark Office rewired how its examiners approach AI eligibility. The change did not come from Congress or the Federal Circuit. It came from a rapid series of internal actions that together add up to the most inventor-friendly posture on software eligibility in years. Anyone with AI applications sitting in prosecution right now has a narrow, valuable window, and letting a case go abandoned this quarter could mean walking away from a claim that has suddenly become grantable.

A reset measured in months

The sequence started on August 4, 2025, when Deputy Commissioner for Patents Charles Kim issued a memorandum to the office’s software-heavy technology centers. The Kim Memo, as practitioners quickly named it, told examiners to stop overusing the “mental process” category. A claim limitation that cannot practically be performed in a human mind, it said, does not belong in that bucket. The memo also reminded examiners that a claim reciting “a particular solution to a problem or a particular way to achieve a desired outcome” is eligible, cautioned against the “Apply It” shortcut of treating a computer as a generic tool, and set a clear evidentiary bar. An examiner should reject under Section 101 only when there is a greater than fifty percent probability that the claim is ineligible. Uncertainty alone, the memo stated, is not enough.

Weeks later the posture hardened into something with precedential weight. On September 18, 2025, the Senate confirmed John Squires as USPTO Director. Squires arrived from private practice with a background in emerging technologies, including AI, blockchain, and fintech, and he moved fast. Eight days after confirmation, on September 26, an Appeal Review Panel he led issued Ex Parte Desjardins. The panel vacated a Section 101 rejection of machine-learning claims, finding that they recited genuine technical improvements in how a model operates. The invention let an AI system learn new tasks in succession while protecting knowledge of earlier tasks, a problem researchers call catastrophic forgetting, and it did so using less storage and reduced system complexity.

Desjardins also gave the profession a quote worth keeping. The panel warned that “categorically excluding AI innovations from patent protection in the United States jeopardizes America’s leadership in this critical emerging technology.” That single sentence reframes eligibility as a competitiveness issue rather than a doctrinal puzzle, and it signals how the current office wants examiners to think. The decision was designated precedential on November 4, 2025, and folded into the Manual of Patent Examining Procedure in December, which means every examiner is now expected to apply it.

The office was not finished. On December 4, 2025, it released a pair of memos on Subject Matter Eligibility Declarations. The first tells examiners to treat a properly submitted declaration as a legitimate route for an applicant to overcome a Section 101 rejection, and it supplies worked examples covering mental processes, technology improvements, and the “significantly more” analysis. The second gives practitioners best practices, urging them to keep eligibility testimony in a standalone declaration so that questions of enablement, written description, novelty, and obviousness stay separate from the eligibility record. For AI and diagnostics, the two technology areas that have absorbed the most eligibility scrutiny, a declaration that documents a concrete technical improvement is now a real tool rather than a long shot.

Read together, these actions carry a clear message for anyone prosecuting AI claims. The door is open, and the burden of proof has shifted. An examiner can no longer wave a claim away as abstract without engaging with the specific technical contribution and clearing a real probability threshold. Applications that stalled under the old reflex deserve a fresh look. A pending case that drew a boilerplate mental-process rejection a year ago may now be worth an amendment, a declaration, or a renewed appeal rather than abandonment.

Why “for now” belongs in the headline

The caution is as clear as the opportunity and it is the reason the title of this piece ends with “for now.” None of this is statute. The Kim Memo, the December declarations guidance, and even the MPEP updates are policy and administrative practice, and a precedential Appeal Review Panel decision reflects the priorities of the current Director. Congress has not amended Section 101. The Alice and Mayo decisions remain good law at the Supreme Court, and the Federal Circuit continues to decide eligibility appeals on its own terms. A future Director could narrow this guidance as fast as the current one expanded it. The reprieve is genuine and the reprieve is fragile, and both facts should shape strategy.

What to do this quarter

The practical takeaway is to use the window without betting the portfolio on it lasting. Reassess pending AI and software applications now, while examiner behavior favors applicants. Draft and amend specifications so they document a specific, measurable technical improvement, the kind of concrete advance that carried Desjardins, because that showing holds up whether eligibility is decided by a friendly memo or by a skeptical court later. Consider a well-built eligibility declaration where the record needs it. Build every claim as though it will one day be tested under a harsher standard than today’s, so that a swing back in policy finds your strongest cases already fortified.

For patent counsel, R&D leaders, technology transfer offices, and novelty consultants, this is not a moment to wait and see. The applications on your desk right now are the ones that benefit from the current posture, and they are the ones exposed if it changes. The office has told the profession what it wants to see. The move that serves clients in both worlds is to give it to them now.

Want more stories like this? Subscribe to the IP.com newsletter for the latest on AI, the patent landscape, and the interesting corners of the IP world worth knowing about.