What’s Behind the Headlines?
The U.S. Patent and Trademark Office (USPTO) has now issued an internal memorandum to examiners, particularly those in software-related technology centers focusing on artificial intelligence (AI) and machine learning (ML). Over the past few years, patent examiners at the USPTO have become increasingly critical of patents incorporating some aspect of AI or ML technology. More and more, examiners have been relying on one particular statute, 35 U.S.C. § 101, to reject these patents. 35 U.S.C. § 101 is the law that defines what is and what is not patentable. This latest instruction to patent examiners reminds them to apply these subject matter eligibility standards more carefully.
Though the memo states it does not establish new rules or procedures, its tone and clarifications signal a shift toward fewer § 101 rejections for AI and software inventions, making it easier to patent innovations in those areas.
Key Takeaways from the Memo
1. Mental-Process Category Narrowed
Examiners are reminded not to classify innovations beyond human capability, such as hardware-based AI operations—as mere mental processes. This is particularly encouraging for AI and ML inventions that rely on computational processes far beyond what the human mind can handle.
2. “Recite” vs. “Involve”: A Crucial Distinction
Now, only claims that explicitly recite abstract ideas trigger a full § 101 analysis. Those that merely involve abstract ideas may be eligible without further scrutiny. This gives applicants more room to frame their patent claims effectively.
3. Technological Improvement Must Be Apparent
Your patent application does not have to spell out technological improvements word for word. Instead, your application must simply describe the invention so that someone skilled in the art sees the improvement clearly, especially when applied to a specific solution rather than applied broadly.
4. Avoid Claiming a Judicial Exception
Courts have ruled that some things, like abstract ideas, fall outside § 101 and are therefore not patentable. The memo highlights the importance of not putting a judicial exception into your patent claims. Something like “training a neural network by…” is fine, whereas something like “training a computer using a propagation algorithm…” requires a mathematical calculation, which is a nonpatentable abstract idea. Remember that generic, high-level training steps without mathematical detail are more likely to be deemed patent-eligible, while training steps that include named algorithms are more likely to be rejected as covering non-patentable subject matter.
5. Raised Bar for § 101 Rejections
Some commentary suggests the memo effectively raises the threshold so that examiners should only issue § 101 rejections when they are clearly warranted. If the examiner is unsure if § 101 applies or if the call is a close one, the memo confirms that the examiner should grant the patent unless they are more than 50% sure that the claim is ineligible under § 101.
What This Means for Inventors and Businesses
- AI, ML, and software developers may see increased success in obtaining patents, provided they craft claims that emphasize concrete, technical contributions.
- Inventors are encouraged to focus on specific, technical improvements—not generalized AI buzzwords.
- While the memo doesn’t change the law, its emphasis on sound analysis and careful claim drafting may help reduce overly aggressive eligibility rejections.
- If you have been considering filing a software or AI patent, know may be the time, before the Patent Office changes its mind again.
Recent Comments