AI Insurance Patents: Who Owns What?
New data from Evident's Insurance AI Patent Tracker shows 166 AI patents filed by 30 major insurers since January 2023. Activity is still 30% below the 2020 peak, but filings are concentrating around a few US carriers. Generative AI is now a major theme, and agentic systems are starting to appear, though still rare. If you work in insurance, this is a signal: the IP race is real, and it's uneven.
"Patents offer a rare window into where insurers are placing their biggest bets on AI," said Alexandra Mousavizadeh, Co-founder and CEO of Evident. "This data shows that innovation is overwhelmingly being driven by a handful of US firms, especially in P&C. While the volume of filings remains modest compared to banking, we're seeing a sharp uptick in generative and agentic capabilities. The IP landscape is shifting from protecting past systems to enabling future ones."
The scoreboard
- Concentration: State Farm (326), USAA (218), and Allstate (136) account for 77% of insurer AI filings tracked over the past decade.
- P&C dominance: 89% of all insurer AI patents come from P&C. Telematics, IoT risk monitoring, and sensor-driven systems help meet "technical contribution" bars in the US and Europe.
- Pace: Patent activity peaked in 2020 and hasn't fully rebounded, despite growing GenAI spend.
- Shift to GenAI: Generative AI grew from 4% to 31% of filings in 2023, focused on customer service and claims.
- Agentic AI: Emerging but limited. Only three insurers have filed agentic patents, with USAA in front.
Why P&C is out in front
P&C has more "hard tech" touchpoints: vehicles, homes, devices, and sensors. Those systems translate into patentable methods and architectures more easily than pure process innovation. If your business has access to telematics, image analytics, or IoT streams, you've got patent fuel. For context on patentability standards, see the USPTO's patent resources and EPO's guidance on AI-related inventions.
Where patents are landing
- USAA: GenAI to clarify aerial imagery for property damage assessment.
- State Farm: ML-driven claims triage and autonomous vehicle fault analysis.
- Allstate: In-vehicle assistant to automate claims and offer behavior-based discounts.
- Swiss Re: Predictive analytics for medical data and anomaly detection.
- MassMutual: Interpretable underwriting and AI-tagged document indexing.
- Liberty Mutual: GenAI-generated release notes for engineering teams.
- Zurich Insurance Group: System to match user-typed addresses to a clean, structured address database.
What this means for your roadmap
- Pick your lane: customer service, claims, underwriting, or risk engineering. File where you'll deploy at scale.
- Prioritize "technical contribution." Tie models to devices, imaging, telemetry, NLP pipelines, or control systems.
- Build with patent claims in mind. Document data flows, model improvements, guardrails, and performance gains.
- Stand up an invention-harvesting cadence with engineering leads and product owners. Monthly reviews, not yearly.
- Use provisional filings to move fast, then iterate. Protect the architecture, not just the model weights.
- Decide patent vs. trade secret early. If it's easily discoverable or reverse-engineerable, patent it.
- Coordinate US/EU strategies. The same idea may need different claim framing across jurisdictions.
- Don't ignore defensive publications to block copycat filings in areas you won't patent.
GenAI and agentic: where to place bets
Generative AI is now a third of filings, mainly for service and claims. That tracks with real impact: faster FNOL, better triage, cleaner documentation, and fewer escalations. Agentic systems are early but worth a watch, especially for multi-step claims handling, subrogation search, and fraud checks. If you test agentic flows, log decision traces and safety controls-they're often core to patentability.
Build an IP operating system for insurance AI
- Create a cross-functional IP council (claims, underwriting, legal, data science). Keep it small and decisive.
- Set invention KPIs: disclosures per quarter, conversion to filings, cycle time from idea to provisional.
- Template everything: problem statement, data sources, model type, evaluation, safety, and infra notes.
- Lock down data rights with partners and vendors. If you can't use or improve on the data, your claims get weaker.
- Pilot, measure, file. Use live metrics (latency, accuracy, loss ratios, leakage) in your claims.
- Track competitor patents. Watch State Farm, USAA, Allstate for signals on where the market is heading.
"The insurance sector is at a crossroads. Either patents remain the domain of a few frontrunners, or they become merely a signal of broader competitive intent. As generative and agentic AI reshape the value chain, insurers will need to decide whether to build IP defensively or lead from the front."
Level up your team
If your underwriting, claims, or product teams need a faster path to GenAI fluency, explore job-specific training here: AI courses by job.
Your membership also unlocks:
