In this dataset it is a filing whose claims tie a software artifact — a package, binary, container image, or similar — to a provenance record such as a signature, attestation, or signed metadata, and then verify, validate or authenticate that record before the artifact is trusted or used. The search covers both title/abstract mentions of provenance verification and claim-level language combining artifact terms with verification mechanisms like SLSA or in-toto. It excludes filings that only discuss general software security without a provenance or origin-verification step.

The ranking covers 60 companies across the 129 records in scope, with the leading filer holding 9 records and the fifth-place holder 6. The five leading filers together account for 28.7% of all 129 records, and the ten leading filers account for 45.7% — meaning roughly half of the field's filing activity sits outside the top ten. That leaves a substantial long tail of organisations with one or a few filings each, which is typical of a technology area still being actively defined rather than one with a single dominant architecture.

Filings rose from 7 in 2017 to a peak of 21 in 2021, then declined to 11 by 2024 — a 48% drop over that three-year window, which is the most recent period the data can treat as complete. Because publication typically lags actual filing by around 18 months, the 2025 and 2026 counts shown in any raw trend line are still incomplete and should not be read as evidence the field is cooling further. The honest read is a strong 2021 peak followed by a real pullback through 2024, with the most recent two years still an open question.

The United States leads with 55 filings, followed by the European Patent Office with 21 and the WIPO PCT route with 13. India (9), Israel (6) and Austria (5) each show smaller but notable filing counts, indicating that while the field is concentrated in US and European jurisdictions, provenance-verification claims are also being pursued through international PCT filings and in a handful of other national offices worth monitoring for regional strategy.

The IPC composition shows G06F and H04L dominating at 72.9% and 48.1% of the 129 records respectively, while AI-model computing (G06N) and pictorial communication (H04N) each sit at only 4.7%. That gap is notable because production machine-learning pipelines increasingly depend on signed model and dataset provenance, yet patent claims specifically targeting AI-model artifact verification are still thin. Filers looking for open claim space should look at model-checkpoint provenance chains and build-to-inference attestation rather than the heavily contested core package/container signing mechanisms.