Daily AI Patent Digest - November 18, 2025
Report Date: November 18, 2025 (00:00 GMT+8 Malaysia Time)
Patents Found: 10 Latest USPTO AI Patents
Collection Period: November 3-17, 2025
Executive Summary
Over the past two weeks, the USPTO has granted 10 significant AI-related patents spanning multiple technological domains. These patents demonstrate continued innovation in generative AI, machine learning, computer vision, and autonomous systems. Major companies including ON24, DeepMind Technologies, Theta Labs, and emerging AI healthcare companies are leading the charge in AI innovation.
Patents by Technology Category
Machine Learning & Deep Learning
1. Theta Labs - Modular Large Language Model (LLM) Guided Tree-of-Thought System
- Patent Number: US 12,462,164
- Publication Date: November 3, 2025
- Filing Date: 2024
- Assignee: Theta Labs
- Abstract: A sophisticated system employing a modular LLM architecture that uses a "Tree-of-Thought" approach to improve problem-solving capabilities. The patent details an innovative method for breaking down complex problems into a tree structure of reasoning steps, enabling LLMs to achieve better accuracy and efficiency in task completion.
- Key Technical Areas: Neural Networks, Tree-of-Thought Reasoning, LLM Architecture
- Social Media Mentions: Strong mentions on Theta Network's X account and LinkedIn, praised by the DePIN community for advancing LLM capabilities
2. DeepMind Technologies - Progressive Neural Networks
- Patent Number: US Patent (issued November 4, 2025)
- Publication Date: November 4, 2025
- Assignee: DeepMind Technologies Limited
- Inventors: Emilien Dupont, Hyun Jik Kim, and others
- Abstract: Describes a progressive neural network architecture that enables machines to learn new tasks while preserving performance on previously learned tasks. This addresses the "catastrophic forgetting" problem in neural networks, a critical challenge in machine learning where models lose previously acquired knowledge when learning new information.
- Key Technical Areas: Neural Networks, Transfer Learning, Continual Learning
- Social Media Mentions: Highlighted by patent law community as exemplifying USPTO's renewed support for AI patent eligibility
Reinforcement Learning & Autonomous Systems
3. Palladyne AI - Closed Loop Tasking and Control of Heterogeneous Sensor Networks
- Patent Number: US 12,452,957 B2
- Publication Date: November 3, 2025
- Assignee: Palladyne AI
- Abstract: A groundbreaking patent for advanced swarming and autonomous system coordination. The system enables real-time closed-loop feedback mechanisms for managing multiple heterogeneous robotic and sensor systems working in coordination. Particularly innovative for cross-domain multi-robot coordination with real-time adaptation capabilities.
- Key Technical Areas: Autonomous Systems, Robotics, Sensor Networks, Real-time Control
- Application Areas: Defense, Military Robotics, Multi-domain Autonomous Coordination
- Social Media Mentions: Featured on military technology and robotics blogs; praised in defense technology circles on LinkedIn
4. Fine-tuning Large Language Model Using Reinforcement Learning
- Patent Number: US 12,437,016
- Publication Date: October-November 2025
- Abstract: Describes methods for fine-tuning LLMs using reinforcement learning with search engine feedback. The innovation enables models to learn directly from user interactions and search results, dynamically improving response quality without full model retraining.
- Key Technical Areas: Reinforcement Learning, LLM Fine-tuning, Feedback Mechanisms
- Application Areas: Search Technology, Content Generation, Information Retrieval
Natural Language Processing & Generative AI
5. Hippocratic AI - Safety-Focused Large Language Model for Healthcare
- Patent Number: US 12,142,371
- Publication Date: November 2025
- Filing Date: 2024
- Assignee: Hippocratic AI
- Abstract: Covers innovations in safety-focused LLMs specifically designed for healthcare applications. The patent details the "Polaris constellation architecture" - a multi-component system ensuring that AI-generated medical guidance meets strict safety standards. Includes mechanisms for bias detection, hallucination prevention, and clinical accuracy verification.
- Key Technical Areas: Natural Language Processing, Safety-Focused AI, Healthcare AI
- Application Areas: Healthcare & Medical, Patient-Facing AI Agents
- Social Media Mentions: Announced following $126M Series C funding (November 3, 2025); celebrated on healthcare tech and AI investment communities
6. Dynamic Pricing of Computational Resources
- Patent Number: US 12,419,202
- Publication Date: September 2025 (signed by USPTO Director John Squires)
- Abstract: An innovative system for dynamically pricing computational resources based on real-time demand and system load. Includes AI algorithms that optimize resource allocation and pricing in cloud computing environments.
- Key Technical Areas: Machine Learning, Resource Optimization, Cloud Computing
- Application Areas: Data Processing & Analytics, Cloud Services, Computational Infrastructure
Computer Vision & Image Recognition
7. VisionWave Holdings - WaveStrike RF Computer Vision Fire-Control System
- Patent Number: US Provisional Patent Application No. 63/892,721
- Filing Date: November 17, 2025
- Assignee: VisionWave Holdings, Inc. (West Hollywood, CA)
- Abstract: Describes an AI-assisted radio frequency (RF) computer vision system for real-time target engagement and fire-control in complex combat environments. The technology converts RF signals into situational awareness and targeting cues using advanced machine learning algorithms, offering enhanced target discrimination in challenging conditions.
- Key Technical Areas: Computer Vision, RF Signal Processing, AI-Driven Targeting
- Application Areas: Robotics & Autonomous Systems, Military/Defense, Autonomous Vehicles
- Status: Provisional Patent (filed November 17, 2025)
- Social Media Mentions: Announced on Investing.com and investor forums; discussed on defense technology channels
Data Processing & Analytics
8. ON24 - AI-Powered 'Key Moments' Technology for Long-Form Video
- Patent Number: US 12,445,698 B2
- Publication Date: November 17, 2025
- Filing Date: Prior
- Assignee: ON24, Inc. (NYSE: ONTF)
- Abstract: A proprietary machine learning system that automatically identifies and extracts "Key Moments" from long-form video content and virtual events. Uses AI-powered behavioral analysis to pinpoint the most engaging segments (product demos, expert insights, Q&A moments), which can be repurposed for personalized campaigns across multiple channels including social media and email.
- Key Technical Areas: Machine Learning, Video Analysis, Content Extraction, Behavioral Analytics
- Application Areas: Content Creation & Media, E-commerce & Retail, Education & Training, Marketing Automation
- Social Media Mentions: Announced on Business Wire (November 17, 2025); strong mentions in marketing technology and B2B engagement communities on LinkedIn
Healthcare & Medical AI
9. Alcon - Machine Learning for Contact Lens Compatibility Prediction
- Patent Number: US 11,944,379
- Publication Date: November 11, 2025
- Assignee: Alcon, Inc.
- Inventors: Kevin Baker and others (Texas-based inventors)
- Abstract: A system utilizing machine learning to predict compatibility of multifocal contact lenses for patients with presbyopia. The patent describes methods for analyzing patient vision data, eye geometry, and lifestyle factors to recommend optimal lens parameters, improving fitting success rates and patient satisfaction.
- Key Technical Areas: Machine Learning, Healthcare Data Analysis, Predictive Modeling
- Application Areas: Healthcare & Medical, Medical Devices, Ophthalmology
- Social Media Mentions: Highlighted in healthcare technology and medical device communities
Social Media & E-commerce
10. PayPal - System for Understanding Influencer Reach in Augmented Media Intelligence
- Patent Number: US Patent (issued November 4, 2025)
- Publication Date: November 4, 2025
- Assignee: PayPal, Inc.
- Inventors: Indian and American inventors (team-based innovation)
- Abstract: Describes a sophisticated AI system for analyzing and quantifying influencer reach within augmented media intelligence ecosystems. The technology uses machine learning to assess content impact, audience engagement metrics, and influencer effectiveness across multiple social platforms and augmented reality contexts.
- Key Technical Areas: Machine Learning, Social Network Analysis, Augmented Reality Intelligence
- Application Areas: E-commerce & Retail, Content Creation & Media, Social Commerce
- Social Media Mentions: Discussed in fintech and social commerce communities; highlighted as part of PayPal's broader AI strategy
