42Maru Has Patented AI Data Verification Methods
The company secured new intellectual property to reduce LLM hallucinations in enterprise reporting applications.
Updated on Sept. 22, 2026 in Artificial Intelligence
Live Poll
Would you trust an AI tool to summarize your company's internal data for important decisions?
42Maru has received patent grants for technologies designed to convert unstructured documents into structured tables and charts. These patented methods aim to improve response accuracy in large language models by verifying outputs against specific company data.
Why it matters
Enterprises struggle with AI hallucinations, where models generate plausible but incorrect information. This technology seeks to address that by creating a verifiable, data-grounded link between LLM outputs and proprietary internal documentation.
The firm now holds 122 patents, including 63 in Korea and 59 across the U.S., Europe, and PCT filings. This total includes 50 new filings made since 2023 as part of an aggressive expansion of their intellectual property portfolio.
The players
42Maru
A Seoul-based AI company specializing in enterprise-grade language models designed to reduce hallucinations.
The details
The patented system operates by ingesting unstructured documents and using a multi-step verification process to check AI-generated answers against an organization's internal source data. By organizing this information into structured formats like tables or charts, the model is designed to increase precision. The company utilizes divisional applications—a legal method to keep an existing patent application open to pursue additional, related claims—to broaden the protection of its core technology.
Timeline
50 patents were filed by the company beginning in 2023.
An interactive communication agent service patent was filed in the U.S. in January 2026.
The latest set of patent filings was announced on September 22, 2026.
The Tech Race
42Maru is positioning its legal portfolio to defend proprietary implementations of Retrieval-Augmented Generation, a framework used to ground AI responses in external data. This move formalizes its competitive stance against general-purpose LLMs by focusing on verifiable enterprise reporting.
Enterprise users can expect these verification mechanisms to appear in 42Maru's language model solutions in staged releases. These updates will primarily affect workflows that rely on accurate document analysis and reporting within the company's existing enterprise platforms.
The takeaway
The company is betting that legal ownership of hallucination-reduction techniques will be a key differentiator in the enterprise market. Watch for upcoming announcements regarding the integration of these features into the company's public sector and international enterprise deployments.
Further reading
For more on how enterprises are refining model accuracy, visit Artificial Intelligence.
Source note: This article includes information reported by Business Korea.
Live Poll
Would you trust an AI tool to summarize your company's internal data for important decisions?
