Cybersecurity

Intel Corporation Multimodal LLM Audio Video Tokenization Patent Application

USPTO published patent application US20260099522A1 by Intel Corporation inventors Kuba Lopatka, Adam Kupryjanow, and Tomasz Szmelczynski for power-efficient tokenization and long-context storage of audio and video data for multimodal large language models. The application covers systems with specialized subsystems for receiving input signals, generating discrete tokens, and buffering tokens for durations ranging from seconds to hours.

Expert Selection System for MoE Large Language Models

USPTO published patent application US20260099697A1 for an expert selection system in Mixture of Experts (MoE) large language models. The system includes a global selector that manages selection between global and local modes, and a pre-fetcher that pre-loads selected global expert sets from main memory into faster cache memory. Inventors include Usman Sajid, Marie Mai Nguyen, Shuyi Pei, Younghoon Kim, and Rekha Pitchumani. The filing date was October 1, 2025, under application number 19347703.

Quantum Circuit Simulation Using Tensor Networks Patent Application

The USPTO published patent application US20260099752A1 on April 9, 2026, covering systems and methods for quantum circuit simulation using tensor networks, invented by Mekena Metcalf. The application addresses quantum kernel methods for classification and demonstrates tensor network effectiveness at scaling this application. The application was filed on October 3, 2024.

Systems and Methods for Privacy-Enabled Biometric Processing

The USPTO published patent application US20260100842A1 filed by Private Identity LLC on May 20, 2025, covering systems and methods for privacy-enabled biometric processing. The invention derives encrypted feature vectors from biometric data and uses deep neural networks to authenticate users while preserving privacy. Homomorphic encryption enables authentication comparisons without decrypting the underlying biometric data.

AI-Based Exam Question Generating Method and System

The USPTO published patent application US20260100140A1 for ViewSonic International Corporation, covering an AI-based exam question generating method and system. The system receives reference materials and exam parameters input by a user, then uses a pre-trained AI model to induce exam question rules and generate compliant exam questions from the reference material. The AI model is pre-trained on educational material datasets to learn question generation patterns.

Multi-Language Model-Based Filtering and Labeling of Datasets

The USPTO published patent application US20260100887A1 assigned to Cisco Technology, Inc. The patent covers methods for labeling telemetry data from computer network entities using a combination of N-gram models and large language models. The labeled data is then used to train predictive maintenance models for network infrastructure. The application was filed on July 28, 2025.

Distributed Training of Compressed Machine Learning Models

USPTO published patent application US20260099761A1 for an apparatus enabling distributed training of compressed machine learning models using quantization-based parameter compression. The invention manages compressed ML model parameters across distributed systems by cycling between compression and decompression states, allowing efficient parameter synchronization over networks. The application was filed on October 7, 2024.

Syntiant Neuromorphic IC Sensor Processing Patent Application

USPTO published patent application US20260100186A1 assigned to Syntiant, covering sensor-processing systems with neuromorphic integrated circuits for keyword spotting. The application was filed December 11, 2025, with inventors Kurt F. Busch, Jeremiah H. Holleman III, Pieter Vorenkamp, Stephen W. Bailey, and David Christopher Garrett. The patent describes systems using neuromorphic ICs with neural networks to extract features from sensor data and arrive at actionable decisions for keyword spotting applications.

Budibase Multiple Critical Vulnerabilities Allow Root Code Execution

CERT-Bund issued security advisory WID-SEC-2026-0990 disclosing multiple critical vulnerabilities in Budibase, an open-source low-code platform. Attackers can exploit these flaws to execute arbitrary code with root privileges, escalate to administrator rights, conduct XSS attacks, manipulate data, cause denial-of-service, or disclose confidential information. CVSS Base Score is 9.6 (critical) and Temporal Score is 8.6 (high). Remote attack is possible.

GitLab CE/EE Multiple Vulnerabilities CVSS 8.5 Allow Information Disclosure and Data Manipulation

CERT-Bund issued security advisory WID-SEC-2026-1010 identifying multiple vulnerabilities in GitLab CE/EE versions below 18.9.5, 18.10.3, and 18.8.9. The vulnerabilities carry a CVSS Base Score of 8.5 (high) and a CVSS Temporal Score of 7.4 (high), with remote attack capability confirmed. An attacker could exploit these flaws to disclose information, manipulate data, bypass security measures, cause denial-of-service conditions, or execute cross-site scripting attacks.