Wwwailaw Corp v. Butler Labs Inc

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Wwwailaw Corp. (Ohio)
  • Defendant: BUTLER LABS INC., dba EVE LEGAL (Delaware)
  • Plaintiff's Counsel: Insight, PLC
• Case Identification: 3:26-cv-05930, U.S. District Court for the Northern District of California (N.D. Cal.), 06/17/2026
• Venue Allegations: Venue is alleged to be proper as Defendant maintains its regular and established principal place of business in the Northern District of California.
• Core Dispute: Plaintiff alleges that Defendant's artificial intelligence platform for legal professionals infringes a patent related to methods for generating long-form documents using large language models.
• Technical Context: The lawsuit operates in the rapidly developing field of generative AI, specifically its application in the legal technology sector to automate complex document drafting tasks.
• Key Procedural History: Plaintiff alleges it provided Defendant with notice of the patent and its infringement via correspondence on May 18, 2026, and June 1, 2026, prior to filing the complaint. This notice may form the basis for allegations of willful infringement.

Case Timeline

Date	Event
2024-10-25	'932 Patent Priority Date
2025-11-04	'932 Patent Issue Date
2026-05-18	Plaintiff sends first notice of infringement to Defendant
2026-06-01	Plaintiff sends follow-up notice of infringement to Defendant
2026-06-17	Complaint Filing Date

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

• Patent Identification: U.S. Patent No. 12,461,932, "Method and System for Transforming Data Using Artificial Intelligence to Generate Content," issued November 4, 2025 (the '932 Patent).

The Invention Explained

• Problem Addressed: The patent's background section describes conventional methods for generating content, such as legal documents, as being resource-wasteful, time-consuming, inefficient, and potentially inaccurate Compl. ¶23 '932 Patent, col. 1:12-25
• The Patented Solution: The invention claims a computer-implemented method to improve AI-based content generation. The process involves receiving unstructured data, transforming it into structured data using a large language model (LLM) and a parsing algorithm, and then using an AI engine to generate a final document ( Compl. ¶24; Compl. ¶25, Compl. ¶abstract). A central feature is a "context-aware feedback loop" designed to overcome the inherent output length limitations of LLMs. This loop detects why generation has stopped (e.g., reaching a token limit) and re-executes the prompt with the partially generated text, enabling the creation of long, cohesive documents that would otherwise be commercially impractical ( Compl. ¶24, Compl. ¶75; '932 Patent, Compl. ¶¶col. 19:8-20). The method also includes executing functions in parallel to reduce processing time Compl. ¶1 '932 Patent, col. 1:40-42
• Technical Importance: The claimed method purports to solve the technical problem of generating commercially viable, long-form documents with generative AI by providing a specific architecture to manage and continue text generation beyond the typical output constraints of a single LLM invocation Compl. ¶1 Compl. ¶75

Key Claims at a Glance

• The complaint asserts independent method Claim 1, independent computer-readable medium Claim 19, and independent system Claim 20 Compl. ¶178
• The essential elements of independent Claim 1 are:
  • (a) receiving unstructured input data;
  • (b) transforming the unstructured input data into structured data using an LLM and a parsing algorithm;
  • (c) generating one or more outputs from the structured data for an AI engine;
  • (d) generating a document by converting the outputs, which includes executing a plurality of functions in parallel;
  • (e) executing a context-aware feedback loop using a prompt that detects the reason generation stopped;
  • (f) providing the completed document to a computing device for presentation when generation stops due to completion; and
  • (g) re-executing the prompt with a partially-generated document string when generation stops due to an output constraint.
• The complaint also asserts numerous dependent claims and reserves the right to assert additional claims Compl. ¶178

III. The Accused Instrumentality

Product Identification

The "Eve Legal Accused System," an AI-powered platform for legal professionals accessible at www.eve.legal Compl. ¶¶16-17

Functionality and Market Context

• The complaint alleges the Accused System is a competing AI legal technology platform that allows legal professionals to "streamline casework" by automating tasks like drafting complaints, discovery responses, and medical chronologies Compl. ¶2 Compl. ¶18 A visual from the Defendant's website, included in the complaint, advertises the ability to "Draft complaints, motions, and more" using the user's templates and case context Compl. p. 8
• The system allegedly uses large language models such as "Claude," "Grok," and "ChatGPT" Compl. ¶18 The core of the infringement allegation is that to produce the multi-page documents it advertises, the Accused System must practice the patented method, particularly the claimed feedback loop for handling LLM output limits Compl. ¶16 Compl. ¶¶81-82

IV. Analysis of Infringement Allegations

'932 Patent Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
(a) receiving unstructured input data;	The Eve Legal system receives unstructured data, such as user-uploaded "templates" and other case documents like communication transcripts and evidence.	¶33; ¶43	col. 1:29-30
(b) transforming the unstructured input data into structured data, wherein the transforming is performed using a large language model and a parsing algorithm configured for the unstructured input data;	The system allegedly transforms uploaded documents into structured data using an LLM and a "parsing algorithm" that chunks large files and provides programmatic access to the information.	¶45; ¶57	col. 1:31-34
(c) generating, based on the structured data, one or more outputs configured to be processed by an artificial intelligence engine;	The system allegedly generates intermediate structured outputs, such as JSON objects containing extracted case details, which are then processed by an AI engine (e.g., Claude Opus 4.5).	¶64; ¶67; ¶68	col. 1:35-38
(d) generating, using the artificial intelligence engine, a document by converting the one or more outputs from a first format to a second format, wherein generating the document includes executing a plurality of functions in parallel to reduce execution time of a processing device	The system allegedly uses an AI engine to generate documents by issuing concurrent LLM invocations for different parts of a document (e.g., individual interrogatory responses) and then assembling them, which constitutes parallel execution to reduce generation time.	¶71; ¶72; ¶75	col. 1:40-42
(e) in response to detecting that generation of the document stopped, executing a context-aware feedback loop using a prompt that detects a reason why the generation of the document stopped;	The system allegedly uses Anthropic's LLM API, which returns a "stop_reason" field. The system inspects this field to determine why generation stopped (e.g., completion vs. length constraint), which forms the basis of a context-aware feedback loop. A visual in the complaint shows Anthropic's documentation for "Handling stop reasons" Compl. p. 16	¶79; ¶84; ¶90	col. 19:8-20
(f) in response to detecting that the generation of the document stopped due to the generation of the document by the large language model being complete, providing the document to a computing device for presentation on a user interface of the computing device; and	When the system detects a completion "stop_reason" (e.g., "end_turn"), it transmits the completed document to the user's computing device for presentation in the user interface.	¶91; ¶92	col. 1:43-44
(g) in response to detecting that the generation of the document stopped due to an output constraint of the large language model, re-executing the prompt with a partially-generated document string.	When the system detects a "stop_reason" indicating an output constraint (e.g., "max_tokens"), it allegedly constructs and re-executes a new prompt containing the partially-generated text to continue the document from where it left off. A code snippet in the complaint illustrates this handling of the "max_tokens" condition Compl. p. 20	¶96; ¶99	col. 36:1-4

• Identified Points of Contention:
  • Scope Questions: The complaint defines "parsing algorithm" broadly as "computer instructions used to obtain programmatic access to relevant sets of information or 'strings'" Compl. ¶45 A potential dispute may arise over whether this term requires a more specific, formal parsing process than what the accused system allegedly performs.
  • Technical Questions: A central technical question is whether the Accused System's process for generating long documents is merely a generic "chunking" or chaining of prompts, or if it specifically performs the claimed steps of (1) detecting the specific reason for a halt in generation by inspecting a "stop_reason" field and (2) re-executing a prompt that includes the prior, partially-generated text. The complaint's reliance on public documentation for the "Claude" API Compl. ¶79 Compl. ¶97 suggests the functionality exists, but demonstrating its implementation in the Accused System will be a key evidentiary burden.

V. Key Claim Terms for Construction

• The Term: "context-aware feedback loop"

• Context and Importance: This term appears to be the core of the asserted novelty, addressing the fundamental problem of LLM output length limits. Its construction will be critical to determining infringement. Practitioners may focus on this term because it distinguishes the claimed invention from simple, sequential prompting.

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The term itself is not explicitly defined in the specification. A party could argue it should be given its plain and ordinary meaning, covering any iterative process where an LLM is re-prompted using context from prior outputs.
  • Evidence for a Narrower Interpretation: The claim language itself requires the loop to use a prompt that "detects a reason why the generation of the document stopped" (e.g., Claim 1(e)). The specification describes a process of detecting a stop, determining the reason (completion vs. output limit), and re-running the prompt with the "full document string" added to the message history, suggesting a specific, multi-step logic '932 Patent, col. 19:30-52

• The Term: "output constraint of the large language model"

• Context and Importance: This term is tied directly to the feedback loop and defines one of the conditions that triggers the re-execution step of Claim 1(g). Its meaning is essential for distinguishing a technical stop from a natural completion.

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: A party might argue this term covers any technical limitation that stops generation, including generic timeouts or network errors.
  • Evidence for a Narrower Interpretation: The specification and the complaint's theory focus specifically on documented LLM API limits, such as reaching a maximum token count '932 Patent, col. 19:46-49 The complaint explicitly alleges that "max_tokens" is such a constraint Compl. ¶98 This suggests a narrower definition tied to predefined API limits rather than general system failures.

VI. Other Allegations

• Indirect Infringement: The complaint alleges inducement of infringement, stating that Defendant "specifically instructed and intended its customers to use the Eve Legal Accused System in such a way that infringes the '932 Patent" Compl. ¶185 This is allegedly supported by instructions on Defendant's website and provided with the system Compl. ¶183
• Willful Infringement: The willfulness allegation is based on alleged pre-suit knowledge. The complaint states that Defendant had knowledge of the '932 Patent "at least since on or about May 18, 2026, when Plaintiff notified Defendant of the '932 Patent and Plaintiff's belief Defendant was infringing" Compl. ¶180 Compl. Exhibit B Continued infringement after this date is alleged to be willful Compl. ¶2

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

1. A central evidentiary question will be one of technical operation: Does the Eve Legal Accused System practice the specific, multi-step logic of the claimed "context-aware feedback loop"? Specifically, does it inspect a formal "stop_reason" field from an LLM API to differentiate between task completion and an "output constraint", and then re-prompt with the partially generated text, as alleged, or does it use a different, non-infringing method to create long-form documents?

2. The case may also turn on a question of claim scope: How will the court construe the term "parsing algorithm"? Will the complaint's broad definition stand, or will the court adopt a narrower construction that requires a more complex process than the alleged "chunking" of files, potentially creating a path to non-infringement for the Defendant?

3. A further issue will be attributable conduct: Since several claim steps involve actions by the end-user's computing device (e.g., presentation on a user interface), the court will need to analyze whether Defendant "directs or controls" the user's system to a degree sufficient to be held liable for direct infringement of the entire method under Federal Circuit precedent, an issue the complaint attempts to address Compl. ¶94