Patenting the Machine’s Mind: Navigating the Uncertainty of § 101 Subject Matter Eligibility in the AI Era
As artificial intelligence balances on the boundary between human innovation and machine output, uncertainty under 35 U.S.C. § 101 continues to test the contours of patentable subject matter. This entry examines how evolving jurisprudence, administrative guidance, and legislative reform collectively seek to balance innovation and legal certainty.
Under 35 U.S.C. § 101, patentable subject matter covers processes, machines, and compositions of matter, while excluding laws of nature, natural phenomena, and abstract ideas.[1] These judicially recognized exceptions have long guided courts and the U.S. Patent and Trademark Office (USPTO) in determining what counts as an invention.[2] As AI technology continues to advance, the line between human and machine creativity has become increasingly difficult to define. This evolution poses complex questions for patent law: whether outputs generated by algorithms can truly be considered inventions, and if such creations fall within the scope of patentable subject matter under § 101.[3]
The Supreme Court’s decisions in Mayo Collaborative Services. v. Prometheus Laboratories, Inc. and Alice Corp. v. CLS Bank International established the traditional two-step test that now governs patent eligibility.[4] First, courts ask whether a claim is directed to a judicial exception (the aforementioned laws of nature, natural phenomena, or abstract idea). Secondly, if the case, courts ask whether the patent claim applies to that judicial exception in a practical way.[5] In practice, this framework has created uncertainty, especially for inventions related to computer-implemented and software-related inventions.[6]
After Mayo and Alice, the Federal Circuit struggled to apply the two-step test consistently, particularly to software-adjacent inventions.[7] A lack of predictability indicates that inventors and examiners have little or confusing guidance on what distinguishes an abstract idea from a patent-eligible technological improvement/invention.[8] While the Alice framework was intended to clarify subject matter eligibility, in practice, “it has only deepened the uncertainty [in modern patent law]”.[9]
Over the past decade, courts have applied the two-step approach in ways that have led to a growing number of invalidations under Section 101— particularly in software, artificial intelligence, and biotechnology.[10] Some professionals in these fields have argued this trend chills the innovation the patent law system is supposed to reward and protect. Conversely, others argue that the courts are utilizing Section 101 and its subsequent common law application to its “logical conclusion in light of rapid technological advancements.”[11] Notably, despite frequent requests to the Supreme Court to provide further clarity to Section 101, the Supreme Court has repeatedly declined to do so.[12] For industries that include software, this absence of further guidance leaves conflicting Federal Circuit interpretations to persist.[13] Patent litigation outcomes now “often hinge on narrow factual distinctions – and sometimes simply on the composition of an appellate panel – making results difficult to predict.”[14] As a result, whether a patent claim survives litigation can affect investment decisions, motivation to innovate, and business models, all of which determine which technologies reach the market and which are abandoned.
This inconsistency is evident in how the Federal Circuit has applied the Alice/Mayo test to computer-implemented inventions. In Enfish, LLC v. Microsoft Corp.,[15] the court upheld a patent on a self-referential database because it improved the way computers operate, emphasizing that not all software claims are inherently abstract.[16] Conversely, in Electric Power Group, LLC v. Alstom S.A.,[17] the Federal Circuit declined to grant patent eligibility, holding that the claim was directed to the abstract idea of collecting and analyzing data.[18] The court also found that it failed to include an inventive concept under the second prong of the Alice/Mayo test.[19] This highlights how similar computer-related inventions can be treated differently depending on how their technological improvement is characterized.[20]
These conflicting precedents (and the aforementioned lack of further guidance) highlight the instability between “abstract ideas” and technological improvements that remain in the context of software-implemented inventions. As artificial intelligence becomes more integrated into modern innovations and inventions, the same uncertainty now applies to AI-generated technologies and inventions.[21]
The uncertainty surrounding § 101 has prompted the USPTO to provide new guidance on how examiners should evaluate AI-related inventions.[22] In August 2025, USPTO Deputy Commissioner for Patents Charles Kim issued a memorandum reminding examiners that claims involving artificial intelligence or computer-implemented technologies should not be categorically rejected under § 101 merely because they rely on algorithms or data processing, and that eligibility must be assessed based on whether the claim, viewed as a whole, integrates the exception into a practical application.[23] Following USPTO director John Squire’s reversal of a Patent Trial and Appeal Board § 101 rejection, where he criticized the panel for applying the Alice/Mayo test too rigidly and warned that overbroad rejections under § 101 “jeopardizes America’s leadership in this critical emerging technology” as it is apparent that “… Examiners and panels should not evaluate [AI] claims at such a high level of generality.”[24] Together, the memo and Director Squire’s decision mark a shift in the USPTO’s approach; urging examiners to take a closer, more grounded look at AI-related inventions rather than issuing broad rejections that could stifle emerging innovation.[25]
As a result of recent USPTO action, Congress is responding to the indicated desire to modernize how patent law is applied to AI and ensure that AI inventions are not categorically excluded via § 101.[26] Congress has attempted to responded to the escalating § 101 uncertainty with the Patent Eligibility Restoration Act 2025.[27] This bill was introduced in both the Senate and House of Representatives, aiming to “explicitly eliminate the so-called judicial exceptions to patent eligibility created by the Supreme Court, which find no support in either the Patent Act or the Constitution.”[28] The bill seeks to remedy the fact that “patent eligibility law… has become confused, inconsistent, and unclear.”[29] If enacted, the bill would “restore patent eligibility to important inventions…” and reset the law of patent eligibility prior to Mayo and Alice.[30]
For companies pouring resources into AI research, clear guidance on § 101 is not just a legal technicality, it changes how they protect, value, and bring innovation to market.[31] Uncertainty around patent eligibility can influence how companies plan their business strategies, discourage investment, and make it harder to negotiate licensing or collaboration deals that depend on proprietary algorithms and data models. As AI continues to expand possibilities in creativity and invention, the patent system needs to keep pace. The patent system keeping pace is critical to not only supporting innovation but also helping the U.S. stay ahead in shaping the next generation of technology. Ultimately, how the USPTO and courts handle § 101 in the context of artificial intelligence will shape whether the patent system drives innovation forward or holds it back in this new era of technology.
NOTES
[1] See 35 U.S.C. § 101; see also Alice Corp. Pty. Ltd. v. CLS Bank Int’l, 573 U.S. 208, 216 (2014) (quoting Ass’n for Molecular Pathology v. Myriad Genetics, Inc., 569 U.S. 576, 589 (2013)).
[2] See, e.g., Manita Rawat, et al., USPTO Issues Memorandum Reminding Examiners Regarding Subject Matter Eligibility Evaluation, Morgan Lewis (Aug. 15, 2025), https://www.morganlewis.com/pubs/2025/08/uspto-issues-memorandum-reminding-examiners-regarding-subject-matter-eligibility-evaluation.
[3] See generally Christina Sperry, Helpful Reminders for Patent Eligibility of AI, Machine Learning, and Other Software-Related Inventions, McCarter & English (Sept. 12, 2025), https://www.mccarter.com/insights/helpful-reminders-for-patent-eligibility-of-ai-machine-learning-and-other-software-related-inventions/ (discussing how AI and other software-related inventions are often rejected by the USPTO, and the USPTO’s subject matter eligibility test).
[4] See Alice, 573 U.S. at 216; Mayo Collaborative Servs. v. Prometheus Labs., Inc., 566 U.S. 66, 71 (2012).
[5] Sperry, supra note 3.
[6] Kurt S. Prange, Developments in Patent Subject Matter Eligibility for Software-Related Inventions, in View of Guvera v. Spotify, Womble Bond Dickinson (May 5, 2025), https://www.womblebonddickinson.com/us/insights/articles-and-briefings/developments-patent-subject-matter-eligibility-software-related (noting that “ever since the U.S. Supreme Court’s Alice decision in 2014, Section 101 has been used to invalidate countless software patents.”)
[7] Maxwell H. Terry, Hello, World? Software Patent Protection Stands Alone Due to Uncertain Subject Matter Eligibility Jurisprudence, 108 Minn L. Rev. 403, (2023) (“While the Supreme Court aimed to develop a consistent standard for distinguishing ineligible and eligible subject matter, this framework has negatively impacted predictability before the USPTO and federal courts.”)
[8] See e.g., Anne Elise Herold Li & Paul Keller, Patent Eligibility Uncertainty Persists after latest Supreme Court Denial, Brownstein (Aug. 15, 2025), https://www.bhfs.com/insight/patent-eligibility-uncertainty-persists-after-latest-supreme-court-denial.
[9] Erik Milch & Elizabeth Shrieves, Patent Practitioners are Unsettled Regarding Seemingly-Settled Section 101 Jurisprudence, JD Supra (Oct. 21, 2025), https://www.jdsupra.com/legalnews/patent-practitioners-are-unsettled-6309761/.
[10] See id.
[11] Id.
[12] See id.; see also Anne Elise Herold Li & Paul Keller, Patent Eligibility Uncertainty Persists after latest Supreme Court Denial, Brownstein (Aug. 15, 2025), https://www.bhfs.com/insight/patent-eligibility-uncertainty-persists-after-latest-supreme-court-denial/.
[13] Id.
[14] Id.
[15] 822 F.3d 1327 (Fed. Cir. 2016).
[16] See Kevin D. Rodkey, Elizabeth D. Ferrill & Youngae Kim, Enfish Hooks a Keeper Under § 101: Software Can Be Non-Abstract, Patent-Eligible Improvement To Computer Technology, Finnegan (May 23, 2016), https://www.finnegan.com/en/insights/blogs/federal-circuit-ip/enfish-hooks-a-keeper-under-%C2%A7-101-software-can-be-non-abstract-patent-eligible-improvement-to-computer-technology.html; see also Amardeep Grewal & Gerard M. Donovan, Federal Circuit’s Enfish is an Important 101 Decision, ReedSmith (May 18, 2016), https://www.reedsmith.com/en/perspectives/2016/05/federal-circuits-enfish-is-an-important-101-decisi.
[17] 830 F.3d 1350 (Fed. Cir. 2016).
[18] Id. at 1354.
[19] See generally, id. (noting that the Federal Circuit found that the claim was not patent eligible because it failed the two prong Alice/Mayo test).
[20] Compare Enfish, LLC, 822 F.3d 1327 (upholding a patent on a self-referential database because it improved the way computers operate) with Elec. Power Grp., LLC, 830 F.3d 1350 (denying a patent because it failed to include an inventive concept).
[21] See Diego F. Freire, USPTO Raises Bar for § 101 Rejections in AI Patents, Dykema (Aug. 14, 2025), https://www.dykema.com/news-insights/uspto-raises-bar-for-101-rejections-in-ai-patents.html (“Just as applying a computer to an abstract idea doesn’t make it patentable, merely incorporating AI, large language models [LLMs], or neural networks into an abstract idea cannot transform it into a patentable application.”).
[22] Memorandum, U.S. Pat. and Trademark Off., Reminders on Evaluating Subject Matter Eligibility of Claims Under 35 U.S.C. 101 (Aug. 4, 2025), https://www.uspto.gov/sites/default/files/documents/memo-101-20250804.pdf; Daniel Rose & Sid Kapoor, USPTO Director Reverses PTAB § 101 Rejection, Reaffirms Patent Eligibility for AI Innovations, Pierson Ferdinand (Oct. 3, 2025), https://pierferd.com/insights/uspto-director-reverses-ptab-101-rejection-reaffirms-patent-eligibility-for-ai-innovations.
[23] Memorandum, U.S. Pat. and Trademark Off., Reminders on Evaluating Subject Matter Eligibility of Claims Under 35 U.S.C. 101 (Aug. 4, 2025), https://www.uspto.gov/sites/default/files/documents/memo-101-20250804.pdf.
[24] Daniel Rose & Sid Kapoor, supra note 22.
[25] See e.g., id.
[26] See e.g., id.
[27] S. 1546, 119th Cong. (2025).
[28] Gene Quinn, Patent Eligibility Reform Returns to the Hill: PERA 2025 Explained, IPWatchDog (May 1, 2025, 3:12 PM), https://ipwatchdog.com/2025/05/01/patent-eligibility-reform-returns-hill-pera-2025-explained/.
[29] Id.
[30] Id. (noting that PERA would eliminate the judicial exceptions to patent eligibility and clarify that AI-related inventions satisfy § 101 when the computer is integral to performing the claimed invention).
[31] See generally Diego F. Freire, USPTO Raises Bar for § 101 Rejections in AI Patents, Dykema (Aug. 14, 2025), https://www.dykema.com/news-insights/uspto-raises-bar-for-101-rejections-in-ai-patents.html? (noting that clarifying § 101 could reduce rejections in borderline AI cases—a welcome development for companies seeking to expand in the AI sector.)