Tag: #ai

NVIDIA confirms $12.9 billion acquisition of Hugging Face

NVIDIA confirmed it will buy Hugging Face for $12.93 billion, saying the AI model hub will remain open and users won't need NVIDIA compute to build.

$NVDA, #nvidia, #huggingface, #ai, #acquisition

NVIDIA says AI outscored top human on IOI 2026 problem set in unofficial run

NVIDIA researchers report an AI scored 535.4 on the IOI 2026 problem set—above the top human's 498.27—but the run was unofficial and unsupervised.

#ai, #nvidia, #competitiveprogramming, #ioi

AMD publishes detailed Instella‑MoE technical report; weights licensed for research use

AMD published a detailed technical report and checkpoints for Instella‑MoE, a 16B MoE model trained on its Instinct GPUs — but model weights are licensed for research only.

$AMD, #amd, #ai, #moe, #openmodels

Judge Rules Pentagon Unlawfully Retaliated Against Anthropic Over Surveillance Stance

A federal judge found the Pentagon unlawfully labeled Anthropic a 'supply chain risk' after the company refused to allow its AI for mass U.S. surveillance.

#anthropic, #ai, #pentagon, #firstamendment

Google rolls out 'agentic video understanding' in Gemini Flash to cut token use and costs

Google adds 'agentic video understanding' to Gemini Flash models to analyze only relevant segments, cutting token use, cost and speeding long-video analysis.

$GOOGL, #google, #gemini, #ai, #video, #cloud

EU Names ChatGPT a 'Very Large' Search Engine; Reddit and Roblox Also Designated Under DSA

EU adds ChatGPT to top-tier DSA search engines, and names Reddit and Roblox very-large platforms—bringing all three under stricter EU oversight.

#ai, #dsa, #chatgpt, #reddit, #roblox

Nvidia reports record $96.2 billion quarter, forecasts $108 billion as AI demand surges

Nvidia posted a record $96.2B quarter and forecast $108B next quarter as data-center AI demand drives revenue; outlook excludes China compute.

$NVDA, #nvidia, #ai, #earnings, #datacenter

LAION publishes 10-million-hour open video dataset LAION-BVD; processed subsets available

LAION published LAION-BVD, a claimed 10-million-hour open video dataset for multimodal AI; processed subsets and pipelines are public, raw videos restricted.

#laion, #dataset, #ai, #video, #multimodal

Apple unveils M6 (first 2nm Mac chip) and quad-die M5 Ultra to boost on‑device AI

Apple introduced the M6 — its first 2nm Mac chip — and the quad-die M5 Ultra, aiming to bring far stronger on-device AI to new Mac mini and Mac Studio.

$AAPL, #apple, #mac, #mseries, #ai

Carnegie Mellon Preprint Finds Reinforcement Learning on Benign Facts Can Amplify LLM Leakage of Memorized Emails

A CMU preprint shows reinforcement learning on benign factual Q&A can make models more likely to reveal memorized Enron email addresses.

#ai, #privacy, #llm, #machine-learning

Conference Paper Finds Many Chatbots Promote Sponsored Options, Often Without Clear Disclosure

Conference paper testing 23 LLMs found many steered users to pricier sponsored options and frequently failed to disclose sponsorship.

#ai, #advertising, #chatbots, #consumerprotection

Nvidia Teams with Six Wall Street Firms to Mobilize Over $500 Billion for AI Infrastructure

Nvidia is partnering with Apollo, BlackRock, Blackstone, Brookfield, Goldman and KKR to create financing platforms aiming to mobilize more than $500 billion for AI infrastructure.

$NVDA, #nvidia, #ai, #finance, #wallstreet

Google Says Gemini App Surpasses 1 Billion Monthly Users

Google announced Aug. 11 that its Gemini app has exceeded 1 billion monthly users, marking rapid growth and deep integration across Google products.

$GOOGL, $GOOG, #google, #gemini, #ai

Study finds structured-output decoding can enable 'control-plane' jailbreaks in LLMs

Researchers find constrained-output grammars can be exploited to jailbreak LLMs; DictAttack hit up to ~99% success on flagship models, paper says.

#ai, #security, #llm, #jailbreak

Preprint finds flaw that let researchers decode encrypted chain-of-thought blobs from major LLM APIs

An arXiv preprint says a cross-model replay flaw let researchers decode encrypted chain-of-thought blobs from major LLM APIs, exposing PII and credentials.

#ai, #security, #privacy, #openai