Meta launched a dedicated Mac app for Meta AI on August 19, 2026, giving the assistant a native desktop home for the first time and pairing it with a set…
The Model’s Tell: Measuring Context-Leakage Attack Signals with Behavior Gauges
arXiv:2608.17829v1 Announce Type: cross Abstract: LLMs increasingly rely on external contexts, such as pre-defined system prompts or retrieved documents,…
Researchers say OpenAI revoked their access to limited cyber program
Multiple cybersecurity researchers said they suddenly lost access to OpenAI’s Trusted Access for Cyber (TAC) program, which offers models with fewer…
BEAR-Bench: A Bilingual Enterprise and Academic Reasoning Benchmark for Multimodal Models
arXiv:2608.17895v1 Announce Type: cross Abstract: While Multimodal Large Language Models (MLLMs) have made significant strides in visual comprehension,…
AI News Brief Hourly Summary 2026-08-19 21h : 14 posts
14 posts published in the last hour 18:32Interpretable Humans, Alien LLMs: Expert Analysis of Latent Structures in Assessment Responses 18:32Meta AI Comes to Mac With Screen Sharing and Business Account Access 18:32Learnware for CSI Feedback: Scene-specific Small Models Can Do…
Interpretable Humans, Alien LLMs: Expert Analysis of Latent Structures in Assessment Responses
arXiv:2608.17810v1 Announce Type: cross Abstract: The evaluation of large language models (LLMs) relies heavily on human-designed assessments, implicitly…
Meta AI Comes to Mac With Screen Sharing and Business Account Access
Meta launched a standalone Mac app for Meta AI on August 19, 2026, giving its chatbot a dedicated desktop home for the first time and turning it toward…
Learnware for CSI Feedback: Scene-specific Small Models Can Do Big
arXiv:2608.17760v1 Announce Type: cross Abstract: Intelligent channel state information (CSI) feedback is essential for realizing the high capacity and…
OpenAI fixes Codex bug that deleted real user files without permission
OpenAI patched Codex after GPT-5.6 Sol started deleting real user files on its own. A cleanup command meant for temporary folders was wiping home…
MotoSafety: Edge-AI with Learned Temporal Importance for Two-Wheeler Collision Risk Assessment Under Time Pressure
arXiv:2608.17823v1 Announce Type: cross Abstract: Powered two-wheeler riders face critical safety challenges in low- and middle-income countries, yet…
Prevalent AI Raises $22M to Scale Trusted Enterprise Context for AI Systems
Prevalent AI has raised $22 million in growth capital from Integrity Growth Partners (IGP), marking the first primary outside investment in the…
What Aggregate Scores Miss: Measuring Item-Level Regressions in Commercial LLM API Migrations
arXiv:2608.17719v1 Announce Type: cross Abstract: Context: Software systems that depend on commercial large language model APIs must migrate to successor…
Waymo Brings Gemini Assistant Into Its Custom Ojai Robotaxis
Waymo riders in the company’s custom-built Ojai robotaxis now have a second AI system in the vehicle with them, and this one is there to talk. Google…
Training with synthetic data for drone detection in thermal imagery
arXiv:2608.17799v1 Announce Type: cross Abstract: Ground-to-Air (G2A) drone detection in medium- and long-wave infrared (MWIR/LWIR) imagery is challenging…
MobileWorldSafety: Benchmarking GUI Agent Safety Against Environmental Injection Attacks in Android Apps
arXiv:2608.17659v1 Announce Type: cross Abstract: LLM-powered GUI agents that autonomously operate smartphones are rapidly transitioning from research…
GADR: Gathering Architecture Decision Records from Meeting Transcriptions
arXiv:2608.17694v1 Announce Type: cross Abstract: Existing LLM-based approaches to Architecture Decision Record (ADR) generation share a critical and…
Communicating Credit Risk with Large Language Models: Evaluation of Explanations from Standard and Alternative Data-Based Models
arXiv:2608.17715v1 Announce Type: cross Abstract: Credit decisioning is a high-stakes task in which model outputs must be accurate and explainable to…
Benchmarking Automated Security Patch Backporting: How Far Are We?
arXiv:2608.17671v1 Announce Type: cross Abstract: Automated security patch backporting is critical for mitigating N-day vulnerabilities. Recent tools…