Home / Research / AI Patent Litigation Statistics

Market research · October 22, 2026

AI Patent Litigation Statistics: 2026 Report

By Darin ten Bruggencate, Founder, AI Industry Reviews

Patents on artificial intelligence, from speech recognition to neural network training, have been asserted in US courts for well over a decade. In July 2026 the University of Tennessee Research Foundation sued Anthropic over Claude Code, one of several recent suits against AI developers. Yet no court-data vendor publishes a free yearly count of AI patent suits, so we built one.

Our analysts matched every granted patent in the USPTO Artificial Intelligence Patent Dataset (AIPD) to the patents listed in the USPTO's district-court litigation files. We found 16,883 cases filed through 2020 that asserted at least one AI patent. The open files end in 2020, so for 2021 through October 2026 we added published counts from Lex Machina, Unified Patents, RPX, the USPTO, LexisNexis Intellectual Property Solutions, Exponent and Docket Navigator, plus court records for major AI cases. In every year from 2015 to 2020, AI patent cases made up between 24.6% and 27.0% of all US patent cases.

In the following sections, we compare AI patent suits by year, published AI-specific counts since 2021, major AI cases and rulings, overall patent filings, plaintiff types, venues, defendants, PTAB challenges and AI patent grants.

AI Patent Suits by Year, 2010 to 2026

Our series through 2020 counts unique district-court cases that asserted at least one granted patent the AIPD labels as AI. No open docket file lists patents in suit after 2020, so later rows carry Lex Machina's all-patent totals and the one published annual AI count.

US District-Court Patent Cases and AI Patent Cases by Year Filed, 2010 to 2026

• AI patent cases peaked at 2,106 filings in 2012 and fell to 878 by 2019. They rose 22.3% in 2020, to 1,074.
• Overall filings dropped to 3,111 in 2023, then climbed 19.6% in 2025, per Lex Machina. The first half of 2026 ran at about 1,900 cases.
• Docket Navigator's 2025 count covers only CPC class G06N, a narrow machine-learning class, while our AIPD series uses the USPTO's broader AI label. Anyone sizing AI patent exposure should keep the two measures separate.

AI Patent Suits Since 2021: What Published Counts Show

No court-data vendor publishes a free annual series of AI patent suits for 2021 to 2026. The published counts below each use a different definition, so we list them side by side.

Published AI Patent Litigation Counts, 2020 to 2026

• Language and vision patents account for most AI litigation events in the LexisNexis set. Generative AI patent families have drawn few suits so far.
• LexisNexis found the top 15 assertion entities hold about 80% of the AI patent families owned by such entities, and the top five about 68%. A small group of licensing firms controls most of the AI patents held for assertion.
• Machine learning events rose more than 2.5 times from 2024 to 2025, which LexisNexis describes as the strongest recent increase of the four domains. For companies that train or deploy models, that is the category to watch in the next vendor update.

Major AI Patent Suits and Rulings, 2021 to 2026

These cases and decisions shaped how AI patents are asserted and defended in the US. We drew them from court opinions, USPTO decisions, dockets and news coverage.

Selected AI Patent Cases and Decisions, 2021 to 2026

• Voice assistants drew the first large AI patent fights, with Alexa facing two separate plaintiffs. AI chips and cloud infrastructure followed in 2024 with suits against Google, Microsoft and Nvidia.
• RPX tracked new AI campaigns in late 2025, including Arlington Technologies on AI avatar and speech systems and Tesseract Systems on neural network training frameworks. The 2026 suits against Anthropic and Eve Legal reach AI application developers.
• Recentive and Desjardins pull eligibility in two directions, with the Federal Circuit rejecting generic machine learning claims and the USPTO accepting claims tied to training improvements. Patent owners and defendants should test AI claims against both decisions before filing or answering.

Overall US Patent Litigation, 2021 to 2026

AI suits are part of the wider patent docket, which three trackers count in different ways. Lex Machina and Unified Patents publish curated case counts, while CourtListener counts dockets marked with the patent nature-of-suit code.

US District-Court Patent Cases Filed, by Tracker, 2021 to 2026

• Lex Machina's 2025 total was the highest of the 2016 to 2025 decade. Unified Patents' count for the same year lands within 16 cases of it.
• Unified Patents counted 18.4% fewer district-court cases in the first half of 2026 than a year earlier. Q3 2026 filings rose to the highest quarterly level of the year.
• Courts awarded $2.353 billion in patent damages across 63 cases in 2025, per Lex Machina, down from $3.48 billion in 2024. Lex Machina counted $8.156 billion in reasonable royalties from 2023 through 2025, against $582 million in lost profits, so royalty-rate evidence drives most damages exposure.

Who Files AI Patent Suits

Patent assertion entities (PAEs) are firms whose main business is licensing and suing on patents. Unified Patents and RPX use the broader term non-practicing entities (NPEs). We labeled each AI patent case's first-listed plaintiff with a name screen: an LLC or Ltd whose name also contains tokens such as IP, Patent, Licensing, Holdings or Technologies. The screen misses assertion entities with ordinary corporate names, so it gives a lower bound.

AI Patent Cases by Plaintiff Name Screen, 2015 to 2020

• Under our name screen, PAE-like plaintiffs filed 15.9% to 22.5% of AI patent cases each year from 2015 through 2020.
• Unified Patents, which classifies plaintiffs by business model, puts NPEs at 55.4% of all patent suits in 2025. NPEs filed 87.7% of high-tech patent suits in Q3 2026.
• For software and device makers, that makes licensing entities the most common source of patent suits.

Defendants Added to US Patent Suits by Plaintiff Type, 2021 to 2025 (RPX)

• RPX counted 2,236 defendants added in NPE suits in 2025, up 44% from 2023. Operating-company litigation grew more slowly over the same period.
• NPE defendant counts in 2025 were just below the 2021 level. Companies facing repeat campaigns can use these counts to benchmark how often they are pulled into suits.

Where AI Patent Cases Are Filed

Counting patent-in-suit rows by district shows where AI patents were asserted most often through 2020. Lex Machina's district counts cover all patent cases from 2022 onward.

Top Districts by AI Patent Assertions (Patent-in-Suit Rows), Cases Through 2020

• The Eastern District of Texas leads with 9,872 AI patent-in-suit rows, almost twice Delaware's total.
• Five of the top six districts are in Texas, Delaware or California. AI patent risk concentrates in a few courts whose judges and local rules shape defense costs.

All Patent Cases Filed in Leading Districts, 2022 to 2025 (Lex Machina)

• W.D. Tex. led in 2022, then fell behind as E.D. Tex. more than doubled its caseload by 2024.
• Lex Machina counted 2,941 E.D. Tex. cases from 2023 through 2025, or 25.7% of all filings, with 2,276 assigned to Judge Rodney Gilstrap. Unified Patents put the district at 34.3% of Q3 2026 filings.
• E.D. Tex. drew more patent cases than any other district in 2025 in both Lex Machina and RPX counts. For an AI company, a new patent suit is more likely to be filed there than in any other court, so defense budgets should assume that court's procedures.

Who Gets Sued on AI Patents

For cases filed from 2015 through 2020 that asserted an AIPD AI patent, we counted defendant rows in the USPTO litigation names file. The 2025 and 2026 lists cover all patent cases.

Most Frequent Defendants in AI Patent Cases, 2015 to 2020

• Microsoft and Apple top the 2015 to 2020 list, with two Samsung entities tied for third.
• Amazon appears under both Amazon Web Services and Amazon.com, so family totals run higher than any single row. Cloud, device and computer companies make up the whole top ten.

Most Frequent US Patent Defendants, 2025 and First Half of 2026

• The 2025 column comes from Lex Machina and the 2026 column from Unified Patents, which groups corporate families. Meta and Microsoft each faced 11 suits in the first half of 2026.
• Apple, Samsung and Google led again in Q3 2026. The device and cloud companies at the top of the AI patent defendant list through 2020 still top the overall patent defendant lists in 2025 and 2026.

PTAB Challenges to Patents, FY 2021 to FY 2026

Defendants often answer a suit by asking the Patent Trial and Appeal Board to cancel the asserted claims through inter partes review (IPR) or post-grant review (PGR), the two trial types created by the America Invents Act (AIA). USPTO statistics do not break out AI patents, so we show the electrical and computer category, where software patents are grouped.

PTAB AIA Trial Petitions and Institution Rates by Fiscal Year

• Petitions fell from 1,433 in FY 2025 to 500 in FY 2026. Electrical and computer patents drew 66.2% of FY 2026 petitions.
• The USPTO denied 438 FY 2026 petitions on discretionary grounds and 114 on the merits. Unified Patents tracked the IPR institution rate falling from 81.8% in January 2025 to 19.4% in August 2025, then reaching 57.3% in Q3 2026.
• Ex parte reexamination requests rose from 437 in 2024 to 726 in 2025, per Unified Patents. AI defendants should budget for reexamination and district-court invalidity defenses when an IPR may not be instituted.

For machine learning patents specifically, Sidley's Alexis Cohen reported in Bloomberg Law that patent owners at least partly prevailed in most PTAB challenges to G06N patents. Most G06N district-court cases ended in settlement, dismissal or invalidity rulings, according to the same analysis.

AI Patent Grants Behind the Litigation

AIPD classifies granted US patents through 2020. The stock of AI patents grew steadily in the decade before our litigation join ends.

Granted US Patents Labeled AI by AIPD, 2010 to 2020

• AIPD counts 668,808 granted AI patents in its 2020 release. Annual grants more than doubled from 2010 to 2020.
• Among 2015 to 2020 AI patent cases, patents with AIPD planning components appear in 5,089 cases and machine learning components in 652.
• Derwent counted more than 32,000 G06N grants in 2025, up from 426 in 2008. Sidley's Alexis Cohen argues in Bloomberg Law that this growing stock of patents points to more AI patent suits in the next several years.

What These Numbers Mean for AI Companies

For companies that build or sell AI products, patent suits so far have come mostly from licensing firms holding older speech, language and vision patents, and they have fallen mostly on large device and cloud companies. The 2026 suits against Anthropic and Eve Legal show the same kind of claim now reaching companies that build AI models and AI applications.

The ways to answer a suit are changing too. Since 2025 the PTAB has instituted far fewer challenges, so defendants have turned to ex parte reexamination and to eligibility motions under Recentive, while patent owners with specific training or model improvements can point to Desjardins when drafting claims.

So if you build or license AI systems, check language, vision and model-training features for patent exposure before launch, and plan for an invalidity defense in district court that does not depend on the PTAB taking the case.