Semantic Engines LLC v. Microsoft Corp

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Semantic Engines LLC (New York)
  • Defendant: Microsoft Corporation (Delaware)
  • Plaintiff's Counsel: Susman Godfrey L.L.P.
• Case Identification: 2:26-cv-00339, E.D. Tex., 04/23/2026
• Venue Allegations: Plaintiff alleges venue is proper in the Eastern District of Texas because Microsoft maintains regular and established physical places of business in the District, including data centers, Microsoft Windows Stores within Best Buy locations, and points-of-presence for its Azure cloud network.
• Core Dispute: Plaintiff alleges that Defendant's Copilot AI assistant infringes three patents related to systems and methods for generating a multi-document summary from search results.
• Technical Context: The technology lies in the field of natural language processing and search, where instead of providing a simple list of links, a system generates a coherent, synthesized digest from multiple source documents to answer a user's query.
• Key Procedural History: The complaint alleges a long history between the parties, starting with outreach from the inventor to Microsoft in 2007 to discuss the technology. It further alleges that the U.S. Patent and Trademark Office cited the asserted patent family as prior art against Microsoft's own AI-related patent applications, and that the inventor again contacted Microsoft in 2025 following these rejections, which may be significant for the Plaintiff's willfulness allegations.

Case Timeline

Date	Event
2007-02-06	Earliest Priority Date for all Asserted Patents
2007-09-01	Plaintiff's inventor allegedly first contacted Microsoft to discuss the technology
2012-08-07	U.S. Patent No. 8,239,358 Issued
2015-12-22	U.S. Patent No. 9,218,414 Issued
2020-09-22	U.S. Patent No. 10,783,192 Issued
2023-02-01	Microsoft launched "Bing Chat" (later rebranded to Copilot)
2025-07-01	Plaintiff's inventor allegedly contacted Microsoft again after USPTO rejections of Microsoft applications
2026-04-23	Complaint Filed

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 8,239,358

• Patent Identification: U.S. Patent No. 8,239,358, "System, Method, and User Interface for a Search Engine Based on Multi-Document Summarization," issued August 7, 2012 ('358 Patent).

The Invention Explained

• Problem Addressed: The patent describes the inefficiency of traditional internet search, where users are presented with a list of links and must manually open and read multiple, sometimes contradictory, sources to find information Compl. ¶¶42-43
• The Patented Solution: The invention proposes a system that goes beyond a list of links to automatically generate a "multi-document summary" Compl. ¶41 It describes a "system core" that passes a query to a "search component," receives results, and sends them to a "summarization component" Compl. ¶85 This summarization component processes the source documents to produce a coherent "digest" composed of extracted sentences and text fragments, complete with reference links back to the original documents '358 Patent, abstract '358 Patent, col. 17:50-65
• Technical Importance: This technology represents a shift from information retrieval (finding links) to information synthesis (providing a direct, consolidated answer), aiming to improve the user experience for complex queries Compl. ¶42

Key Claims at a Glance

• The complaint asserts at least independent Claim 1 Compl. ¶85
• The essential elements of Claim 1 include:
  • sending a query to a system core, which passes it to a search component;
  • the system core receiving results and passing them to a summarization component;
  • the summarization component processing the documents to produce a multi-document summary by analyzing syntax and semantics, extracting sentences/fragments with attached reference links, and combining them into a digest;
  • the system core receiving and displaying the multi-document summary;
  • displaying the original search results in addition to the summary;
  • enabling user interaction to refine the query, navigate the summary, and view source documents, among other functions.
• The complaint reserves the right to assert additional claims.

U.S. Patent No. 9,218,414

• Patent Identification: U.S. Patent No. 9,218,414, "System, Method, and User Interface for a Search Engine Based on Multi-Document Summarization," issued December 22, 2015 ('414 Patent).

The Invention Explained

• Problem Addressed: As a continuation-in-part of the '358 Patent, the '414 Patent addresses the same fundamental problem of inefficient search Compl. ¶¶42-43
• The Patented Solution: The '414 Patent extends the core invention to multimedia content. The claims explicitly require the system to handle "a plurality of media formats" and for the summarization component to process "non-textual media elements" '414 Patent, Claim 1 The specification notes that this can help a user's understanding with "visuals, audio, and other media objects" '414 Patent, col. 25:56-62
• Technical Importance: This update acknowledges the evolution of the web beyond text-only content and aims to provide synthesized answers that incorporate relevant images, videos, and other media Compl. ¶122

Key Claims at a Glance

• The complaint asserts at least independent Claim 1 Compl. ¶115
• The essential elements of Claim 1 are largely parallel to Claim 1 of the '358 Patent, with the key additions of:
  • The searched documents comprising a "plurality of media formats";
  • The summarization component being configured to "generally process said related documents including non-textual media elements."
• The complaint reserves the right to assert additional claims.

Multi-Patent Capsule: U.S. Patent No. 10,783,192

• Patent Identification: U.S. Patent No. 10,783,192, "System, Method, and User Interface for a Search Engine Based on Multi-Document Summarization," issued September 22, 2020 ('192 Patent).
• Technology Synopsis: As a further continuation-in-part, the '192 Patent refines the invention by adding limitations related to modern user interfaces and data processing. The claims specify query input via "at least one of a voice input and a touch input" and add an explicit step of "removing duplicate results from said multi-document summary" '192 Patent, Claim 1 '192 Patent, abstract
• Asserted Claims: The complaint asserts at least independent Claim 1 Compl. ¶149
• Accused Features: The complaint alleges that Copilot is configured to accept voice and touch inputs and that it handles duplicate information from multiple web sources when constructing its summary Compl. ¶152 Compl. ¶158

III. The Accused Instrumentality

• Product Identification: The accused functionality is Microsoft Copilot, an AI assistant previously known as "Bing Chat" Compl. ¶¶45, 50
• Functionality and Market Context:
  • Microsoft Copilot is an AI tool that utilizes a proprietary technology called "Prometheus," which "combines the 'Bing search index, ranking, and answer results with the creative reasoning capabilities of OpenAI's most-advanced GPT models'" Compl. ¶51 A "Bing Orchestrator" component manages the workflow between user queries, Bing Search, and the Generative AI (GenAI) model to produce a conversational, summary-style answer Compl. ¶¶51, 87 The complaint alleges this functionality is available as a standalone product and is integrated into Microsoft's Edge browser, Windows 11, and Microsoft 365 suite Compl. ¶¶47-48 A visual in the complaint shows Copilot's integration across various Microsoft applications like Word, Excel, and Teams Compl. p. 19
  • The complaint positions Copilot as central to Microsoft's business and AI strategy, citing metrics such as over 100 million monthly active users and 15 million paid enterprise seats Compl. ¶¶52-53

IV. Analysis of Infringement Allegations

'358 Patent Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
sending a query to a system core where said query is passed to a search component for searching the documents...	A user submits a query to Copilot, which is sent to the "Prometheus and/or Bing Orchestrator (the 'system core')," which then passes the query to Bing Search (the "search component").	¶88	col. 21:36-43
said system core in turn receiving results from said search component... and passing to a summarization component a specified number of said results...	The system core receives search results from Bing Search and passes them to Microsoft's GenAI component (the "summarization component").	¶88	col. 21:55-62
where said summarization component processes related documents... to produce a multi-document summary comprising a digest... by analyzing syntax and semantics... extracting sentences and text fragments... with an attached reference link... and combining said sentences and text fragments... into said digest...	The GenAI component, utilizing a large language model, analyzes the search results to generate a "digest." This generated response includes extracted or synthesized text with attached reference links to the source documents.	¶¶91-93	col. 22:1-9
said system core receiving said multi-document summary from said summarization component;	The Prometheus/Bing Orchestrator system core receives the completed summary from the GenAI component.	¶94	col. 22:45-46
receiving from said system core said multi-document summary for display as a product of said query;	Copilot displays the multi-document summary to the user as the primary answer.	¶95	col. 22:47-48
steps for enabling the user to use the presentation... to refine the query...	Copilot enables users to ask follow-up questions to refine or modify the query and results. The complaint provides a screenshot of suggested follow-up topics presented to the user Compl. p. 38	¶98	col. 22:50-67

'414 Patent Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
said searched documents comprising a plurality of media formats...	Bing Search is alleged to search documents that "comprise a plurality of media formats, including text and images and/or video."	¶118	col. 25:35-37
wherein said summarization component is configured to generally process said related documents including non-textual media elements... to substantially produce a multi-document summary...	The GenAI component is alleged to be configured to process related documents that include non-textual media elements to produce a summary. A screenshot shows Copilot generating a summary about Abraham Lincoln that includes images Compl. p. 45	¶121; ¶122	col. 25:40-49
extracting sentences and text fragments of said related documents...	The GenAI component extracts or synthesizes sentences and text fragments from the related source documents to form the summary digest.	¶123	col. 26:1-6
enabling said user to use a presentation... to choose a language for said query and said related documents...	Copilot allows the user to switch languages for the query and the resulting summary. The complaint provides a screenshot of Copilot responding to a request to switch to Spanish Compl. p. 47	¶132	col. 26:42-44

• Identified Points of Contention:
  • Technical Question: A central issue may be whether the "summarization component" of the patents, which is described as extracting and combining pre-existing "sentences and text fragments," can be read to cover a modern generative AI model. A generative model synthesizes new sentences based on its training and the source material, rather than strictly extracting them. The complaint alleges Copilot "extracts sentences and text fragments" Compl. ¶93, but a defendant may argue this mischaracterizes how an LLM functions.
  • Scope Question: For the '414 Patent, a point of contention may be whether merely placing an image from a source document into a generated summary, as shown in the complaint's visual evidence Compl. p. 45, satisfies the claim limitation of "process[ing] said related documents including non-textual media elements" to "produce a multi-document summary."

V. Key Claim Terms for Construction

• The Term: "multi-document summary"

  • Context and Importance: This term is the core of the invention. Its construction will be critical in determining whether a modern, LLM-generated conversational response, which synthesizes information rather than just extracting it, falls within the patent's scope.
  • Intrinsic Evidence for a Broader Interpretation: The specification refers to the summary as a "digest" that involves "extracting and blending together significant and relevant aspects of the results" '358 Patent, col. 17:29-32 The term "blending" may support an interpretation that goes beyond simple extraction and combination.
  • Intrinsic Evidence for a Narrower Interpretation: The claims of all three patents consistently recite "extracting sentences and text fragments... from which each has been extracted" and "combining said sentences and text fragments" '358 Patent, Claim 1 '414 Patent, Claim 1 '192 Patent, Claim 1 This language strongly suggests an extractive process, where the output is composed of verbatim pieces of the source documents, which could support a narrower definition that excludes generative synthesis.

• The Term: "summarization component"

  • Context and Importance: Practitioners may focus on this term because its definition will determine whether Microsoft's "GenAI component" (an LLM) is equivalent to the claimed component. The dispute may turn on whether the patent covers only the extractive summarization technology of its time or is broad enough to include modern generative AI.
  • Intrinsic evidence for a Broader Interpretation: The claims define the component by its function: it "processes related documents" to "produce a multi-document summary" '358 Patent, Claim 1 This high-level functional language could be argued to be technology-agnostic.
  • Intrinsic Evidence for a Narrower Interpretation: The detailed description explains that the engine "parses each individual document to extract key semantic concepts" and "constructs a summary comprising sentences or text fragments taken out of different documents" '358 Patent, col. 18:27-33 This detailed description of an extractive process may be used to argue for a narrower construction that excludes generative models.

VI. Other Allegations

• Indirect Infringement: The complaint alleges active inducement of infringement under 35 U.S.C. § 271(b). The basis for this allegation is Microsoft's creation and dissemination of extensive promotional materials, user guides, training videos, and a "Scenario Library" that allegedly instruct and encourage customers and end-users to use Copilot in a manner that directly infringes the asserted patents Compl. ¶¶106-107 The complaint includes a screenshot of a Microsoft resource titled "Identify initial high-value Copilot scenarios" as evidence of this instruction Compl. p. 23
• Willful Infringement: The complaint alleges willful infringement based on Microsoft's purported knowledge of the patents. The allegations point to pre-suit knowledge dating to at least September 2007, when the inventor allegedly contacted Microsoft to discuss the technology Compl. ¶¶65-66 The complaint further alleges that Microsoft had knowledge through its own patent prosecution, where the USPTO cited the asserted patent family as prior art against Microsoft's applications, and that the inventor re-engaged with Microsoft in 2025 after these events, only to be rebuffed Compl. ¶¶70-80

VII. Analyst's Conclusion: Key Questions for the Case

• A core issue will be one of technological scope: can the claims, which describe an extractive summarization process (i.e., "extracting" and "combining" text fragments), be construed to cover Microsoft's Copilot, which employs a generative AI model (LLM) to synthesize new text based on source documents? The case may turn on the distinction between extracting existing content versus generating new content.
• A central claim construction question will be the definition of a "multi-document summary." The court will need to decide if this term is limited to a collection of directly extracted text fragments with links, as the patent language suggests, or if it is broad enough to encompass a synthesized, conversational answer generated by an LLM that cites its sources.
• A key question for damages will be willfulness: given the detailed allegations of a nearly two-decade history, including direct communications in 2007 and Microsoft's alleged knowledge of the patents through its own patent prosecution efforts, the court will have to resolve whether any infringement by Microsoft was willful and deliberate.