USPTO ASAP Pilot Gives Insurance Patent Filers a New Edge

The USPTO extended its Artificial Intelligence Search Automated Pilot Program on April 16, 2026, doubling the intake target, waiving the petition fee, and pushing the window to June 1. ASAP runs an AI prior art search before any examiner opens the file. For insurance carriers, where three filers hold most of the AI patents, that changes when overlap becomes visible. The extension also exists because almost nobody enrolled.

Key Takeaways

• 169 petitions filed and 76 granted as of April 16, 2026, which is 4.8% of the expanded 3,200-application target and about 10.6% of the original 1,600 goal.
• The $450 petition fee is waived for filings on or after March 23, 2026, and the per-Technology-Center cap doubled from 200 to 400.
• Up to ten ranked documents arrive in an Automated Search Results Notice that carries no response obligation and is not a notification under 35 U.S.C. Section 132.
• 77% of insurer AI patents sit with State Farm at 326, USAA at 218, and Allstate at 136, so the prior art an ASRN surfaces for one is frequently a patent held by another.
• The tool searches the specification, not the claims, in the version applicants get through June 1.

What the Extension Changed and What It Revealed

ASAP subjects eligible applications to an automated prior art search before substantive examination. The USPTO tool takes the CPC designation, specification, claims, and abstract as context and searches US patents, US pre-grant publications, and the Foreign Image and Text database.

The output, an Automated Search Results Notice, lists up to ten documents ranked by AI-determined relevance plus a search string applicants can run in Patent Public Search. It requires no response and creates no presumption that cited references must be addressed. It does enter the file wrapper at publication, and examiners treat its documents like any others in the search file.

The uptake figure is the news inside the extension. As of April 16, only 169 petitions had been filed across all Technology Centers, with 76 granted, against a target that now stands at 3,200. Set against roughly 650,000 utility applications a year, even the full target is under 0.5% of volume.

Three things suppressed participation. The original October 2025 launch required the petition on the same day as the application, excluding anyone who learned of the program afterward. The $450 fee added friction to an unproven benefit. And practitioners reported that the references came back tangentially relevant, a complaint that traces directly to a design limitation covered below.

Eligibility also excludes continuing applications. Insurance portfolios built on continuation chains, where child applications refine parent claims, cannot enroll the part of the portfolio that often carries the value.

The Economics Depend Entirely on Filing Volume

The value of ASAP is the timing, not the notice, which is advisory. An applicant normally waits 12 to 24 months for a first office action before learning which references troubled the examiner. ASAP moves that signal ahead of examination, making a preliminary amendment possible before the first action is written.

That is where the money is. An applicant who narrows claims to clear the flagged references can convert what would have been a rejection into an allowance, removing a round of office actions that costs $3,000 to $8,000 in attorney time on a typical insurance AI prosecution. It also lands before the 12-month foreign filing decision, where the alternative is committing $50,000 to $100,000 per jurisdiction without knowing whether the prior art supports broad claims.

Filing volume decides whether any of that matters. State Farm has filed 326 AI-related patents since 2014, USAA 218, and Allstate 136, together 77% of all insurer AI patents on Evident's tracker. State Farm's total implies roughly 30 to 40 new filings a year, so one fewer office action per application compounds into a real budget line. A carrier filing two or three AI applications a year carries the same ASRN review overhead against a fraction of the benefit.

Concentration also determines what the notices will say. When three filers dominate telematics risk scoring and claims automation, the closest prior art to a new State Farm application is often an existing USAA or Allstate patent. ASAP surfaces those intra-industry collisions before prosecution rather than during it, which turns a first-action surprise into a filing-stage decision about whether to narrow around a competitor or move the claim elsewhere.

The newer claim categories are the exception. Generative AI grew from 4% to 31% of insurer AI filings between 2014 and late 2025, and USAA's multi-agent coordination filings sit in a genuinely thin field. There, an ASRN confirming sparse coverage is a reason to claim more broadly, not less.

It Does Not Touch the Rejection That Actually Blocks These Claims

ASAP works on Sections 102 and 103, novelty and obviousness. It does nothing for Section 101, and Section 101 is the dominant rejection ground for insurance AI applications.

The two failures need different medicine. A prior art rejection is cured by narrowing to distinguish over cited references, which is exactly what early visibility accelerates. An abstract-idea rejection says the claims recite a method of organizing human activity or a mathematical concept without an inventive concept, and no quantity of prior art changes that reading. A claim to using machine learning to predict claim severity draws the second kind regardless of how novel the implementation is.

So the program helps applications that were already going to get through. Filings with specific model architectures and defined data transformation pipelines benefit, because prior art is their actual bottleneck. Broadly functional claims hit the Section 101 wall before prior art becomes relevant, and the Ex parte Desjardins decision, the Kim Memo, and the Subject Matter Eligibility Declaration procedure that reshaped that landscape operate on a separate track from ASAP.

The tool has a second limit that explains the complaints. Director Squires disclosed that the current version searches the specification only, not the claims. Prior art relevance is assessed against claims, so a search reading specification context can surface references tied to disclosed but unclaimed embodiments while missing what matters to the actual scope. A Version 2 searching claims directly is described as forthcoming, and applicants filing through June 1 will not receive updated results retroactively.

There is also a disclosure question the waiver does not resolve. The Office says the notice needs no response and no Information Disclosure Statement. But under 37 C.F.R. Section 1.56, counsel who has read a material reference owes a duty of candor whether or not the notice compels it, as Nixon Peabody flagged. A free search that generates known references a team must then evaluate is not free, and for a portfolio in active enforcement territory the conservative answer is to submit everything, which puts the cost back where the fee waiver took it out.