State Farm, USAA, and Allstate Hold 77% of Insurer AI Patents: What the Concentration Means for the Rest of the Industry

From tracking USPTO filings across the top 30 insurers over the past 18 months, the three-carrier concentration pattern has been consistent. State Farm, USAA, and Allstate have dominated insurer AI patent activity since 2014, and the gap between these three and everyone else keeps widening. But the shift toward agentic AI filings in late 2025 marks a new strategic phase worth examining in detail, because the rest of the industry’s patent gap may matter more than their AI press releases suggest.

Evident, the AI benchmarking and intelligence platform for financial services, released its Insurance AI Patent Tracker in December 2025, quantifying what many in the industry suspected but had never seen measured: three P&C carriers account for more than three-quarters of every AI patent filed by an insurer over the past decade. The data covers 30 major insurers across North America and Europe and reveals 166 AI patents filed since January 2023 alone. This piece analyzes what the concentration means for competitive moats, which patent categories each carrier favors, and why the freedom-to-operate question is becoming urgent for everyone else.

The numbers: a 77% concentration ratio

The headline figure is striking. Since 2014, State Farm has filed 326 AI-related patents. USAA has filed 218. Allstate has filed 136. Together, those 680 patents represent 77% of all AI patents filed by insurers tracked by Evident. The remaining 23% is distributed across 27 other carriers, many of them household names with significant technology budgets.

Insurer	AI Patents (Since 2014)	Share of Total	Primary Patent Focus
State Farm	326	~37%	Claims triage, autonomous vehicle fault analysis
USAA	218	~25%	Aerial imagery GenAI, agentic multi-agent systems
Allstate	136	~15%	In-vehicle AI assistant, telematics, behavior-based pricing
All other insurers (27)	~203	~23%	Various: predictive analytics, document processing, risk modeling

Property and casualty carriers dominate overall, holding 89% of all insurer AI patents. This structural advantage reflects P&C’s natural overlap with technologies that produce patentable claims: telematics sensors, IoT-driven risk monitoring, image recognition for damage assessment, and real-time pricing models that respond to sensor data. Life and health carriers, where AI applications tend to involve less hardware integration and more actuarial model refinement, have filed comparatively few patents.

“Patents offer a rare window into where insurers are placing their biggest bets on AI,” said Alexandra Mousavizadeh, cofounder and CEO of Evident. “This data shows that innovation is overwhelmingly being driven by a handful of U.S. firms, especially in P&C.”

What each carrier is patenting

The aggregate concentration number obscures meaningful differences in what each carrier chooses to protect. Patterns we’ve seen across the three portfolios reveal distinct strategic bets.

State Farm: claims automation and autonomous vehicles

State Farm’s 326-patent portfolio, the largest in the industry, clusters heavily around two pillars: machine learning for claims triage and AI systems for autonomous vehicle fault analysis. The claims triage patents cover methods for automatically sorting incoming claims by complexity and urgency, routing the straightforward ones to straight-through processing while flagging complex or suspicious cases for human review. These are operationally significant patents because they describe the core workflow that determines how quickly (and cheaply) a carrier can process its claims inventory.

The autonomous vehicle patents represent a longer-term strategic bet. State Farm has filed extensively on systems that analyze sensor data from autonomous and semi-autonomous vehicles in real time, detect collision events, determine fault allocation, and generate claim information automatically. As connected and autonomous vehicles increase as a share of the insured fleet, these patents position State Farm to own key methods in the machine-to-insurer data pipeline. The commercial value depends on how quickly Level 3 and Level 4 autonomy reaches mainstream adoption, but the IP foundation is being laid now.

State Farm also holds patents across user mobility profiling, processing graph computation for insurance workflows, and AI-driven 3D landscape modeling using LIDAR data. The breadth of the portfolio suggests a deliberate strategy to patent across the full claims lifecycle, from first notice of loss through settlement.

USAA: generative AI and the agentic frontier

USAA’s 218-patent portfolio is notable less for its size than for its forward positioning. USAA leads among all insurers in agentic AI patents, a category so new that only three insurers have filed in it at all. Agentic AI systems go beyond generating content or predictions; they take autonomous action, coordinate multiple AI agents, and operate within feedback loops that allow them to refine their behavior without human intervention at each step.

USAA’s most visible patent application in the generative AI space involves using GenAI to clarify aerial imagery for property damage assessment. After a hurricane or severe storm, adjusters historically relied on manual inspection or raw satellite imagery to estimate damage. USAA’s patented approach uses generative models to enhance, clarify, and annotate aerial imagery, speeding up the catastrophe claims cycle. For a carrier that serves military families often stationed far from their insured properties, this technology has clear operational value.

The agentic patents are more architecturally ambitious. They describe multi-agent coordination systems for underwriting and claims workflows, where specialized AI agents handle discrete tasks (document extraction, risk scoring, compliance checking) and pass structured outputs to one another through orchestration layers with continuous feedback mechanisms. This continues a trend we’ve documented in our coverage of AIG’s agentic underwriting architecture, but USAA’s filings emphasize coordination patterns and adaptive feedback loops rather than single-pipeline processing.

USAA’s broader AI strategy operates under what the company has described publicly as an “experiment and see” approach. The carrier ran multiple generative AI pilots in 2023, implemented roughly a dozen production solutions in 2024, and uses a multi-platform strategy spanning Google Gemini, AWS Bedrock, open-weight models like Meta’s Llama, and internally trained models. The patent portfolio reflects this pragmatic approach: protecting specific innovations that emerge from production deployments rather than filing speculative claims.

Allstate: telematics and in-vehicle intelligence

Allstate’s 136-patent portfolio emphasizes the intersection of AI and telematics, a natural extension of its Drivewise usage-based insurance program. The most distinctive patent application describes an in-vehicle AI assistant that automates elements of the claims process and offers behavior-based discounts in real time. This system uses vehicle sensors to monitor driving behavior, categorize the driver’s accident risk profile, and dynamically adjust pricing signals.

Allstate has also filed on telematics-based driving assessment technology that processes sensor data in real time to evaluate driving quality. The Drivewise program already measures mileage, braking patterns, speed, and time-of-day driving patterns to calculate discounts. The patents extend this into AI-driven categorization and predictive risk scoring, moving beyond simple actuarial rating variables toward continuous behavioral monitoring.

Additional Allstate patents cover interpretable AI for underwriting decisions and AI-tagged document indexing, reflecting the same operational efficiency goals seen in the other two portfolios but with a distinctive emphasis on explainability. This focus on interpretable AI is notable because it aligns with the NAIC’s emerging regulatory framework that may require carriers to explain AI-driven decisions to regulators and consumers.

The generative AI surge: from 4% to 31% of filings

One of the most important trends in the Evident data is the rapid growth of generative AI patents as a share of total insurer AI filings. Between 2014 and the early 2020s, generative AI represented roughly 4% of patent applications. By October 2025, that share had climbed to 31%. The increase reflects the industry’s operational priorities: generative models are being deployed primarily in customer service automation and claims processing, the two highest-volume, highest-cost functions in most carriers’ operations.

The concentration of generative AI patents in claims and customer service is consistent with where carriers report the fastest ROI. Customer service chatbots, automated first-notice-of-loss intake, claims summarization for adjusters, and policyholder communication drafting are all areas where generative models can reduce cost per transaction quickly. Morgan Stanley’s recent projection that AI will cut P&C expense ratios by 200 basis points, generating $9.3 billion in operating income by 2030, is driven largely by these same use cases.

The broader patent data supports this focus. Claims and underwriting together account for over 300 patents across the industry, more than twice the volume of the next largest category (customer service and risk modeling). Risk pricing patents exist but represent a smaller share, possibly because pricing model innovations are harder to patent (they often rely on well-known statistical techniques applied to new data) and easier to protect as trade secrets.

Agentic AI: the next strategic frontier

If generative AI patents reflect where insurers are deploying today, agentic AI patents signal where they plan to compete tomorrow. As of the Evident tracker’s December 2025 release, only three insurers had filed agentic AI patents, with USAA leading the category.

Agentic AI differs from generative AI in a fundamental way. Generative models produce outputs (text, images, structured data) in response to prompts. Agentic systems take autonomous action: they execute multi-step workflows, coordinate with other AI agents, make decisions within defined guardrails, and adjust their behavior based on outcomes. In insurance, the natural applications include end-to-end claims processing (from intake through adjudication and payment), complex underwriting workflows where multiple data sources must be assessed sequentially, and regulatory compliance monitoring that responds to filing deadlines and rule changes without human scheduling.

Evident expects agentic patent activity to increase in 2026, with a growing focus on system-level designs, multi-agent coordination, control mechanisms, and continuous feedback loops as insurers formalize successful agentic use cases into patentable architectures. The high-impact areas for these filings are underwriting and claims handling, where the complexity of real-world workflows creates room for genuinely novel system designs.

The scarcity of agentic patents is itself a signal. Most carriers are still in the experimentation phase with agentic systems, running pilots and proofs of concept rather than filing IP claims. By the time they formalize their architectures, the early filers (USAA in particular) will have prior art and issued claims that constrain how latecomers can design their own systems. This is the IP moat forming in real time.

The 2020 peak and the patent decline puzzle

One of the counterintuitive findings in the Evident data is that insurer AI patent activity peaked in 2020 and remains roughly 30% below that high-water mark, even as carrier investment in AI, and particularly in generative AI, has accelerated. Several explanations account for this gap.

Trade secret preference. The most compelling explanation is strategic. For many AI innovations, trade secret protection offers advantages that patents do not. Patents require public disclosure of the invention, which gives competitors a blueprint even as it provides legal exclusivity. Trade secrets require no disclosure and can protect implementations indefinitely, as long as the secret is maintained. With the passage of the Defend Trade Secrets Act (DTSA) in 2016 providing a federal cause of action, trade secret litigation has become more practical. Trade secret filings increased 25% within a year of the DTSA’s passage, and over 1,200 trade secret cases were filed in U.S. courts in 2025 alone. For carriers deploying AI models whose competitive value depends on specific training data, feature engineering, and hyperparameter tuning, trade secret protection may be more practical than patenting the underlying method.

Section 101 uncertainty. The legal landscape for AI patent eligibility has been in flux. The Federal Circuit’s April 2025 decision in Recentive Analytics, Inc. v. Fox Corp., which gave courts a template for invalidating generic “apply ML to new data” claims, followed by the Supreme Court’s December 2025 cert denial, created a chilling effect on broad AI patent filings. We covered this in detail in our analysis of the USPTO Section 101 reset. Carriers and their patent counsel are aware that broadly drafted AI claims face significant invalidity risk, which may discourage speculative filing even as specific, technically detailed applications remain viable.

Vendor delegation. A growing share of AI implementation at carriers runs on vendor platforms rather than internally developed systems. When a carrier deploys Guidewire’s ClaimCenter with AI-assisted triage, or uses Verisk’s Synergy Studio for catastrophe modeling, the patentable innovation belongs to the vendor, not the carrier. The rise of AI platform vendors in insurance has effectively shifted patent activity from carrier R&D departments to the vendor ecosystem. This redistribution of innovation does not show up in the Evident tracker, which focuses on insurer-filed patents.

Open-source substitution. The rapid maturation of open-source AI tools and foundation models (LLama, Mistral, and others) has reduced the need for carriers to develop patentable proprietary methods for common tasks. If an open-source model can handle claims summarization or document classification adequately, there is less incentive to invest in a patent-worthy alternative.

The long tail: what the other 27 carriers are filing

Beyond the top three, the Evident data reveals a long tail of low-volume filers. Swiss Re has patents covering predictive analytics for medical data and anomaly detection systems. MassMutual has filed on interpretable underwriting models and AI-tagged document indexing. Liberty Mutual and Zurich have various specialized AI applications in their portfolios. But none of these carriers comes close to the filing volume of the top three.

The 166 patents filed since January 2023 by all 30 tracked insurers represent a modest pace, roughly 5.5 patents per insurer over a two-year period. Remove the top three carriers’ recent contributions and the per-insurer average drops further. For context, AIG alone has three significant AI underwriting patents that we’ve analyzed in depth across our patent cluster, and EXL, a services company rather than a carrier, holds 10 AI patents spanning document extraction, knowledge graphs, and a domain-specific insurance LLM. The vendor-side patent portfolio often dwarfs the carrier-side portfolio at mid-market companies.

Freedom-to-operate risk for mid-market carriers

The concentration of AI patents among three carriers creates specific risks for the rest of the industry. These risks fall into three categories.

1. Licensing exposure

As patent holders become more sophisticated about monetizing their portfolios, carriers that independently develop AI systems may discover they infringe on methods already claimed by State Farm, USAA, or Allstate. Machine learning-driven claims triage, for example, is a capability that dozens of carriers are building or buying. If State Farm’s patents cover key aspects of how those triage systems work, any carrier using a similar approach could face licensing demands or, in the worst case, infringement litigation.

The practical risk is modulated by the Recentive Analytics precedent, which weakens the enforceability of broadly drafted AI claims. Narrowly drafted claims tied to specific technical implementations remain enforceable, and those are exactly the kind of patents that sophisticated filers like State Farm tend to hold. Our Section 101 analysis details how the post-2025 eligibility framework treats different claim types.

2. Build-vs-buy complications

Mid-market carriers evaluating whether to build AI capabilities internally or purchase them from vendors face an additional variable: does the planned system overlap with patented methods? A carrier building its own ML-driven claims triage system needs to consider whether its architecture infringes State Farm’s claims, while a carrier purchasing a vendor solution inherits whatever freedom-to-operate analysis (or lack thereof) the vendor has performed.

This is the dynamic we explored in our analysis of the AIG vs. Quantiphi patent race, where the build-vs-buy decision is complicated by overlapping IP claims from both carriers and vendors. The three-carrier concentration adds another layer: even if a vendor’s platform is clear of vendor-owned patents, the carrier deploying it may still infringe on patents held by one of the top three filers.

3. Competitive moat effects

The deepest long-term impact may be competitive rather than legal. Carriers with large patent portfolios can deploy patented methods freely while their competitors must either license, design around, or accept the risk of infringement. Over time, this creates a compounding advantage: the patent-holding carrier iterates faster because it does not need to navigate its own IP restrictions, while competitors face higher development costs and longer timelines as they work to avoid infringement.

Whether this theoretical moat becomes a practical one depends on enforcement posture. To date, none of the top three insurer patent holders has launched significant patent enforcement campaigns against other carriers. But the portfolios exist, and defensive use (cross-licensing in disputes, deterring new entrants) may be as strategically valuable as offensive assertion.

Why P&C dominates: a structural advantage

The 89% P&C share of insurer AI patents is not simply a function of the top three carriers being P&C companies. It reflects structural characteristics of the P&C business that make AI innovations more patentable.

P&C insurance generates rich, heterogeneous data from physical sources: telematics devices, IoT sensors, aerial and satellite imagery, vehicle diagnostic systems, weather stations, and property inspection tools. AI systems that process this data often involve specific hardware integrations, novel signal-processing methods, or unique combinations of sensor inputs. These technical implementations produce claims that are more likely to survive Section 101 scrutiny than the purely algorithmic approaches common in life and health insurance AI.

Additionally, P&C’s short-tail lines offer faster feedback loops. An AI claims triage system deployed on auto physical damage can be evaluated for accuracy within weeks; a mortality prediction model deployed in life underwriting may take years to validate. Faster feedback means faster iteration, more refinements, and more occasions to develop patentable improvements to the initial system.

The telematics ecosystem provides an additional patent surface. Usage-based insurance programs like Allstate’s Drivewise, Progressive’s Snapshot, and State Farm’s Drive Safe & Save all involve AI-driven analysis of driving behavior data collected from mobile devices or vehicle sensors. Each step in the data pipeline, from collection through analysis to pricing integration, offers opportunities for patentable innovation that simply do not exist in life or health insurance workflows.

What the patent gap reveals about AI strategy

The disconnect between AI press releases and AI patent filings is one of the most telling features of the Evident data. Nearly every large insurer has announced AI initiatives, hired chief AI officers, partnered with technology vendors, and published thought leadership about their digital transformation journeys. Yet the patent record shows that only a handful are developing proprietary AI innovations significant enough to protect.

This gap has several interpretations. The charitable reading is that many carriers are wisely choosing trade secret protection over patents, particularly for innovations involving model training and data engineering. The less charitable reading is that many carriers’ AI programs are primarily deployment of vendor technology rather than genuine innovation, and there is nothing proprietary to patent because the intellectual property belongs to the platform provider.

Mousavizadeh framed the question directly: “Either patents remain the domain of a few frontrunners, or they become merely a signal of broader competitive intent.” The evidence currently favors the first interpretation. The top three carriers are not just filing more patents; they are filing in emerging categories (agentic AI, autonomous vehicle systems) that the rest of the industry has barely begun to explore.

Implications for actuarial practice

The concentration of AI patents among three carriers has specific implications for practicing actuaries.

Pricing and underwriting actuaries at carriers outside the top three should be aware that AI-driven methods they develop or deploy may overlap with patented approaches. This does not mean actuaries need to become patent attorneys, but it does mean that AI governance frameworks at carriers should include freedom-to-operate reviews as a standard step before deploying new AI-driven pricing or underwriting models. ASOP No. 56 (Modeling) already requires actuaries to understand the models they rely on; understanding the IP landscape around those models is a natural extension.

Reserving actuaries should note that the claims automation patents held by the top three carriers describe methods that affect claim settlement timing and amounts. If your carrier adopts similar methods, the resulting changes in settlement patterns will flow through loss development triangles and may require assumption updates. Predictive analytics in underwriting already creates data drift that actuaries must account for; AI-automated claims processing adds another dimension of operational change to monitor.

Consulting actuaries advising mid-market carriers on AI adoption should factor IP risk into their recommendations. A carrier that builds a claims triage system similar to State Farm’s patented approach faces different risks than one that licenses a vendor platform. The build-vs-buy analysis should incorporate not just development cost and time-to-market, but also freedom-to-operate assessment relative to the 680+ patents held by the top three filers.

Looking ahead: what to watch in 2026

Several developments will shape how the patent concentration story evolves over the next 12 months.

Agentic patent acceleration. Evident expects agentic patent filings to increase as carriers move from experimentation to production deployment. USAA’s early lead in this category gives it a structural advantage, but the field is new enough that other carriers and vendors can still establish meaningful positions. The insurers that file first on multi-agent coordination architectures for specific insurance workflows will shape the design space for everyone who follows.

Post-Recentive filing strategy. The Recentive Analytics decision and the November 2025 USPTO guidance changes are already affecting how carriers draft AI patent claims. Expect to see more technically specific claims tied to particular hardware integrations, data processing pipelines, and actuarial calculation methods, and fewer broad “AI applied to insurance” claims. The carriers that adapt their filing strategies to the new eligibility framework will build more durable portfolios.

Vendor-carrier IP overlap. As AI platform vendors like Guidewire, Verisk, Duck Creek, and Earnix expand their own AI patent portfolios, the interaction between vendor patents and carrier patents will become more complex. Carriers need to understand not just whether their internally developed systems infringe on other carriers’ patents, but also how their vendor agreements allocate IP rights for AI innovations developed on vendor platforms. The EXL patent portfolio analysis illustrates how a services company can accumulate significant insurance AI IP that overlaps with carrier implementations.

Regulatory influence on patent value. The NAIC’s ongoing work on AI regulation, including the potential transition from bulletin to enforceable model law, could affect the practical value of AI patents. If regulators require carriers to use explainable AI methods, patents on black-box approaches lose operational relevance even if they remain legally valid. Conversely, patents on interpretable AI methods (like some of Allstate’s filings) could become more valuable as regulatory requirements increase.

Why this matters

The insurance industry has a long history of technology adoption following a concentrated-then-diffused pattern: a few innovators develop proprietary methods, the methods prove their value, and then vendors commoditize them for the broader market. Mainframe-era policy administration systems, early catastrophe models, and telematics platforms all followed this arc.

AI may follow the same pattern, but with an important difference. Earlier waves of insurance technology innovation occurred before the current patent landscape took shape. The carriers that built the first cat models did not patent them aggressively; the methods diffused through the vendor ecosystem relatively freely. Today’s AI innovators are patenting aggressively, creating legal barriers that did not exist in earlier technology cycles.

The question is whether those barriers hold. The Recentive Analytics decision weakened the enforceability of some AI patent claims, but technically specific patents remain strong. The 680 patents held by State Farm, USAA, and Allstate represent a formidable collection of specific, production-tested AI methods. Whether those patents function as competitive moats, licensing revenue streams, or merely defensive shields will depend on how aggressively the holders choose to enforce them.

For the 27 other carriers in Evident’s tracker, and for the hundreds of smaller carriers that were not tracked at all, the message is clear: the AI patent landscape is not democratizing. Innovation leadership is consolidating, and the cost of catching up increases with every filing cycle.