Microsoft Research has announced VibeVoice-ASR-Streaming-1.5B, a streaming ASR model that continuously transcribes who said what in real-time as audio is input.
Google has announced a multi-year partnership with MrBeast. In a video released on September 5, Gemini provides advice for wilderness survival.
Microsoft Research has released VibeVoice-ASR-Streaming, a streaming speech recognition model capable of continuously transcribing who said what in real-time as audio is input.
Google has announced Gemini 3.8 Flash, which features enhanced reasoning and coding capabilities, and Gemini 3.8 Flash Cyber, designed specifically for cybersecurity.
Google announced on 2026-09-03 the launch of the Fairwind Program, which provides advanced AI-driven cyber defense capabilities to government agencies and trusted partners.
Google has released Gemini 3.8 Flash as the next-generation model in the Gemini 3 family, featuring performance enhancements in software engineering and agentic knowledge workflows.
US-based Anthropic has released a free web page that can read metadata embedded in image, audio, and video files to detect traces of processing by the AI assistant Claude.
On September 3, 2026, OpenAI announced its new frontier model, GPT-6 Astra.
Multiple lawsuits have been filed against OpenAI and CEO Sam Altman, alleging they provided assistance to a suspect in a school shooting that occurred in Tumbler Ridge, Canada.
NYC Mayor Mamdani has announced a policy banning the use of AI in classrooms for approximately 600,000 public school students in 8th grade and below for one year, starting from the 2026-27 academic year.
WebLLM is an engine that utilizes WebGPU for hardware acceleration, enabling the direct execution of language model inference within the browser without the need for a server.
An investigation into Perplexity's responses across 380 software categories revealed that many of the cited domains are new sites outside of major rankings.
A verification using 310 factual questions regarding 210 technology companies investigated the accuracy of numerical citations in Perplexity's search models, revealing that approximately 30% of citations are inaccurate.
IBM and Confluent have announced the early access availability of time series foundation models on Confluent Cloud, enabling real-time forecasting and anomaly detection by embedding models directly into data streaming.
AISLE's AI system discovered vulnerabilities corresponding to 6 CVEs in the curl communication library, where AI systems from OpenAI and Anthropic failed to find any.
A Hacker News contributor pointed out three challenges regarding the comparison graphs of LLM intelligence and cost published by ArtificialAnalysis.
A research team from Binghamton University has shown through brain imaging analysis that in aging brains, the hippocampus may blend different memories together instead of faithfully reproducing individual ones.
While the technological evolution of humanoid robots is remarkable, reports suggest that challenges such as physical manipulation and general-purpose utility remain for practical labor applications.
Mistral has announced that in certain cases, excluding the enterprise tier, user input and output data such as conversations and documents may be included in its model training programs.
Anthropic announced the launch of a tool that allows users to verify whether a file was created or processed using Claude, utilizing signatures based on the C2PA standard within metadata.