Fastest
Eureka on the web
When you want the answer in the next five minutes.
The agent works the prompt against patents and technical literature, citing every source.
Run your analysis now →https://www.patsnap.com/resources/blog/rd-blog/synthetic-data-generation-patent-landscape/ · Patsnap · data cut-off 2026-07-31 · downloaded from the live page
Patent Landscape · Synthetic Data GenerationSynthetic Data Generation Patents: Who Holds the Claims and Where They Thin Out
- 25.8% concentration at the top. The five most active filers combined hold 158 of 613 records in scope, but the remaining share spreads across a long tail of single- and few-filing entrants.
- Filings grew 64% from 2021 to 2024. The count rose from 56 to 92 records over that span, the last window unaffected by publication lag.
- G06N dominates, B25J and H04L barely register. AI-model computing classes cover 51.5% of records while robotics and digital-transmission overlaps sit under 8.5%, marking where the technology has and hasn't been claimed.
613
Published Records
26%
Top-5 Share of All Records
+64%
Filing Growth 2021→2024
US
Leading Jurisdiction
Filing growth compares 2021 (56 records) with 2024 (92) — a three-year span. 2024 is the most recent year we treat as complete: publication lags filing by roughly 18 months, so 2025 onwards are still filling in and any growth rate that ends there would understate the field. Top-5 share is the combined record count of the five largest assignees divided by all 613 records in scope (CR5), not by the ranked leaders only.
Published byPatsnap Research··6 min readSourced from Patsnap Eureka
Overview
What the synthetic data generation patent record shows
Synthetic data generation covers methods for producing artificial datasets that stand in for real-world data during model training or testing — spanning adversarial generation, domain randomization, conditional sampling and techniques aimed at rare-event coverage or privacy leakage control. The 613 records in scope span filings from 2015 through the current data cut-off, with the underlying search string tying generation methods to specific fidelity and privacy concerns rather than generic data-augmentation claims.
Because publication lags filing by roughly 18 months, the most recent one to two years in any trend understate real filing activity; 2024 is the last year that can be read as materially complete.
Filing activity and technology composition, 2015–2026
Source: Patsnap Eureka. Assignee ranking and totals. Derived from a Patsnap search on Synthetic Data Generation covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
The Data
Filing trend and technology composition
Two views of the same 613-record dataset: filing activity by year, and the IPC subclasses those records claim into.
Filing trend, 2017–2026
Filings rose from zero recorded activity in 2017 to a peak of 154 in 2025, with growth of 64% between 2021 (56) and 2024 (92) — the last window not distorted by publication lag. 2025 and 2026 figures will continue to fill in as later publications post.
IPC subclass composition
G06N (AI-model computing) appears in 51.5% of the 613 records, well ahead of G06T image processing (26.8%) and G06F digital data processing (24.3%). Robotics (B25J, 8.0%) and digital transmission (H04L, 6.5%) show the technology reaching into adjacent domains at much lower density. Because records can carry multiple IPC classes, these shares sum to more than 100%.
Shares are the percentage of the 613 records in scope. A patent can carry several IPC classes, so the shares add up to more than 100%.
Source: Patsnap Eureka. Filing trend and technology composition. Derived from a Patsnap search on Synthetic Data Generation covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.
Go deeper on Synthetic Data Generation with Eureka
This page is one run against one query. Ask Eureka your own question about synthetic data generation and every answer comes back with the patent numbers behind it.
Try EurekaKey Patents
Most-cited records in the dataset
Representative Recent Filing
Post-generation input stream modification for inferred network-based simulated data (WO2026170185A1)
CITIBANK, N.A.
The filing describes a system that receives modification requests for simulated datasets carrying updated privacy parameters, retrieves inferred entity-relationship networks capturing structural, semantic and statistical properties of the simulated data, and selectively regenerates only the data records affected by a privacy parameter change rather than the entire dataset.Filed by CITIBANK, N.A.; published 2026-08-13.
Highest-citation records
| # | Publication no. | Patent title | Citations |
|---|---|---|---|
| 1 | US20190295302A1 | Segmentation Guided Image Generation With Adversarial Networks | 392 |
| 2 | US20200285939A1 | Aggressive development with cooperative generators | 129 |
| 3 | US20200160178A1 | Learning to generate synthetic datasets for traning neural networks | 112 |
| 4 | US20190095798A1 | Stochastic categorical autoencoder network | 105 |
| 5 | US20230067841A1 | Image Enhancement via Iterative Refinement based on Machine Learning Models | 102 |
| 6 | US20250175456A1 | Ai-controlled sensor network for threat mapping and characterization and risk adjusted response | 94 |
| 7 | US20210383538A1 | Multi-modal, multi-resolution deep learning neural networks for segmentation, outcomes prediction and longitu… | 90 |
| 8 | US20230095092A1 | Denoising diffusion generative adversarial networks | 88 |
| 9 | WO2020028382A1 | Multi-modal, multi-resolution deep learning neural networks for segmentation, outcomes prediction and longitu… | 86 |
| 10 | US10839269B1 | System for fast and accurate visual domain adaptation | 83 |
Citation counts favour older records simply by virtue of being searchable longer; treat them as a signal of influence within this corpus, not as a ranking of current technical importance.
Each row carries its publication number; clicking a row searches Eureka by that number.
Source: Patsnap Eureka. Citation counts and representative records. Derived from a Patsnap search on Synthetic Data Generation covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Run it yourself
Put your own technology through the same analysis
Where to run it
For builders
MCP server & REST API
When it has to run inside your own pipeline.
Patent search, landscape analysis and assignee resolution as MCP tools. Drop them into any agent framework, or call REST directly.
Browse MCP servers →Insights
What the numbers mean for a filing strategy
Three read-throughs from the concentration, growth and classification figures above.
Concentration
25.8%
of 613 records held by top 5 combined
The top of the field is thin relative to the whole
158 of 613 records sit with the five most active assignees, leaving nearly three-quarters of filings spread across a long tail. That pattern suggests the field is still open to new entrants rather than locked up by a handful of blocking portfolios.
Top 5 combined vs. all 613 records
Growth
+64%
growth 2021 to 2024
Filing activity accelerated through the last complete window
The rise from 56 records in 2021 to 92 in 2024 shows sustained investment in the years before publication lag starts to mask true volume. Read 2025's headline peak of 154 with that lag in mind rather than as a slowdown signal.
2021: 56 records → 2024: 92 records
Classification
51.5%
of records carry G06N
Claims cluster on the AI-model layer, not the application layer
Over half of records touch G06N (AI-model computing), while healthcare informatics (G16H, 9.1%) and robotics (B25J, 8.0%) show far lower density. That gap points to underclaimed application-specific synthetic data methods even where the core generative techniques are heavily filed.
G06N 316 records vs. B25J 49 records, of 613 total
Eureka AI Agent
Looking for what nobody has claimed yet?
Eureka can read the same corpus for gaps instead of for coverage: under-claimed branches adjacent to synthetic data generation, with the prior art for and against each one.
Source: Patsnap Eureka. Co-assignee relationships and derived observations. Derived from a Patsnap search on Synthetic Data Generation covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Players
Who is filing, and where momentum sits
The ranked leaders reflect a mix of large AI infrastructure firms, life-science and healthcare researchers, and telecom and automotive entrants working on adjacent applications.
Leader
51
records
A single filer well ahead of the field
The leading assignee holds 51 records, more than double the fifth-place filer's 22, giving it a clear lead within the ranking without approaching majority share of the 613 records in scope.
Leader vs. fifth place: 51 vs. 22
Mid-field
22 → 11
fifth to tenth place
A steep drop-off after the top five
Record counts fall from 22 at fifth place to 11 at tenth, and the top 10 combined account for 37.4% of all 613 records — meaning most of the ranked leaders hold single-digit or low-double-digit counts.
Top 10 combined: 229 of 613 records
Momentum
-88% to -50%
YoY among recent filers
Recent-year filing counts are down across the tracked assignees
Every assignee with recent-year momentum data shows a year-over-year decline, from -50% to -100%. Given the roughly 18-month publication lag, this looks more like incomplete recent-year data than an actual pullback in R&D activity.
Latest-year filings, individual assignees
Recent-year filing momentum
| Assignee | Recent year | YoY |
|---|---|---|
| D5AI LLC | 4 | -88% |
| NVIDIA Corp | 1 | -86% |
| Google LLC | 1 | -50% |
| Memorial Sloan Kettering Cancer Center | 1 | 0% |
| D5AI LLC | 0 | -100% |
| Regeneron Pharmaceuticals Inc | 0 | — |
| Robert Bosch GmbH | 0 | -100% |
| ATOMBEAM TECH INC | 0 | -100% |
Source: Patsnap Eureka. Assignee-level momentum. Derived from a Patsnap search on Synthetic Data Generation covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
What's Next
Where to take this next
The dataset points to specific next steps depending on whether the goal is freedom-to-operate, portfolio strategy, or spotting a filing gap.
Check freedom-to-operate against the most-cited records
The highest-citation filings in this dataset, particularly around adversarial generation and synthetic dataset construction for neural network training, are worth clearing before committing engineering resources to similar methods.
Explore citing records in EurekaMap the under-claimed branches to a first draft claim
Rare-event coverage sampling and privacy-parameter-driven regeneration show low filing density relative to core generative methods, which is where a narrowly scoped first claim is more likely to clear prior art.
Draft claim language in EurekaTrack the ranked leaders' recent activity, not just their totals
Cumulative counts favour early movers; watching year-over-year filings for the leading assignees, with the publication-lag caveat in mind, gives a better read on where competitive investment is currently heading.
Set up assignee monitoring in EurekaSource: Patsnap Eureka. Forward-looking reading of the same dataset. Derived from a Patsnap search on Synthetic Data Generation covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
FAQ
Common questions about synthetic data generation patents
The dataset's ranking is led by a single assignee with 51 records, well ahead of the fifth-place filer at 22. The top five combined hold 158 of 613 records in scope, or 25.8%, while the top ten hold 37.4%. That leaves the majority of records spread across a long tail of filers with single-digit counts, so no one company controls the field outright.
Filings grew 64% between 2021 (56 records) and 2024 (92 records), the last window that can be read as materially complete. Later years, including the reported 2025 peak of 154, will continue to be revised upward as more filings publish, since publication typically lags filing by about 18 months. Read any apparent recent-year decline with that lag in mind rather than as a genuine slowdown.
G06N, covering AI-model-based computing, appears in 51.5% of the 613 records in scope, making it the dominant classification. G06T (image data processing, 26.8%) and G06F (digital data processing, 24.3%) follow. Because a single record can carry multiple IPC classes, these percentages add up to more than 100% and should not be read as mutually exclusive shares of the dataset.
The classification data shows much lower filing density in application-specific areas such as healthcare informatics (G16H, 9.1%), robotics (B25J, 8.0%) and digital transmission (H04L, 6.5%) compared with the core AI-model layer. Sub-areas like rare-event coverage sampling, privacy-parameter-driven regeneration and domain randomization for robotics show visibly thinner claim coverage relative to the core generative methods, making them reasonable starting points for a freedom-to-operate search before drafting.
WO2026170185A1, filed by CITIBANK, N.A. and published 2026-08-13, covers post-generation modification of simulated datasets in response to updated privacy parameters. Its core mechanism uses an inferred entity-relationship network to identify which specific data records are affected by a privacy change, then selectively regenerates only those affected portions rather than the whole dataset. Anyone building selective-regeneration or privacy-parameter-update features for synthetic data should review its claim scope closely before implementing a similar approach.
Answers are grounded in the same dataset. Derived from a Patsnap search on Synthetic Data Generation covering 2015–2026, data cut-off 2026-07-31. Counts reflect published records only and shift as new filings publish.Run this in Eureka MCP
Research Synthetic Data Generation in depth with Eureka
Go past this page: query the whole synthetic data generation corpus yourself, in your own scope.
Every answer comes back with patent numbers you can open.
Disclaimer. This page is generated from Patsnap Eureka data drawn from a limited snapshot of global patent and scientific-literature records, and is provided for general information and reference only.
Patent data carries inherent limitations: recent filings (typically the most recent 18–24 months) are under-counted due to standard publication lag; counts may be reported at either a patent-family or a patent-record basis and are not always directly comparable; classification, applicant-name, and citation data may contain errors, duplicates, or omissions; and the underlying search query defines and constrains the scope shown. As a result, the analysis may be incomplete or inaccurate and may not reflect the full technology landscape.
Nothing on this page constitutes an exhaustive prior-art, novelty, freedom-to-operate, or validity search, nor does it constitute legal, financial, investment, or professional advice, and it should not be relied upon as such. Any patent, commercial, or strategic decision should be verified independently and reviewed with qualified patent, legal, and domain professionals. Patsnap makes no warranties, express or implied, as to the accuracy, completeness, or fitness for any particular purpose of the information presented.
Machine translation. Assignee and organisation names originally recorded in Chinese, Japanese or Korean have been rendered into English by an AI translation step so that the tables stay readable. These renderings are best-effort and may not match a company's registered English name; the original name is what the underlying patent record carries, and it is what any Eureka query launched from this page uses.