Technology News
545 articles
AI technology breakthroughs, research papers, patents, and technical innovations from leading AI startups.

Non-engineer develops multi-million-yen image & video AI tool using Anthropic's Claude Code
A Japanese journalist with no professional programming background used Anthropic's Claude Code to build a sophisticated image and video generation web UI called "百夜スタジオ" (Hyakuya Studio) from scratch....
Elon Musk admitted in court that xAI used distillation to extract knowledge from OpenAI models for t...
Elon Musk admitted in court that xAI used distillation to extract knowledge from OpenAI models for training Grok. The admission came during Musk's testimony in his lawsuit against OpenAI, where he see...
Nvidia's push into physical AI lifts Asian suppliers SK Hynix, Samsung, LG Electronics
Nvidia's expanded physical AI strategy — encompassing robotics, autonomous systems, and AI-enabled manufacturing — has triggered a rally across its Asian supply chain. LG Electronics jumped 15% on rep...

Xiaomi has released MiMo-V2.5-Pro, a 1-trillion-parameter open-source foundation model, claiming it...
Xiaomi has released MiMo-V2.5-Pro, a 1-trillion-parameter open-source foundation model, claiming it approaches the performance of Claude Opus 4.6 on long-horizon agent tasks. The model is the latest i...
NVIDIA launches HealDA, an AI-based data assimilation model for global weather analysis
NVIDIA has introduced HealDA, a data assimilation model built on the HEALPix grid within its PhysicsNeMo framework. The model is designed to integrate sparse observational data into global weather and...
Hugging Face BigCode models leaderboard hosts DeepSeek Coder generations
The BigCode project's models leaderboard on Hugging Face now includes community-submitted generation outputs from DeepSeek's open-weight code model, DeepSeek Coder 6.7B Instruct. The files, contribute...

OpenAI launches GPT-5.4-Cyber variant for trusted cyber defense program
OpenAI announced the launch of GPT-5.4-Cyber, a fine-tuned variant of its GPT-5.4 model designed specifically for defensive cybersecurity use cases. The model is being made available through the compa...

Okta Japan Publishes 'Businesses at Work 2026' Report Revealing 11x Surge in AI Agent-Driven Access Requests
Okta Japan released its annual 'Businesses at Work 2026' report, analyzing data from the Okta Integration Network (OIN) covering November 2024 to October 2025. The report highlights a dramatic increas...
Nvidia releases Nemotron 3 Nano Omni, first in Nemotron series to natively support audio alongside text, images, and video inputs.
The model launch represents Nvidia's latest push into compact multimodal AI, a strategy that positions its chip ecosystem to capture inference workloads at the edge. By releasing the model openly on H...

Anthropic clashes with White House over expansion of 'Mythos' AI security system
Anthropic has hit a roadblock in expanding access to its advanced security AI system 'Mythos', as the White House pushed back on plans to extend availability from roughly 50 organizations to 120. The...

티오리, 리눅스 커널 취약점 'Copy Fail' 발견
South Korean cybersecurity startup 티오리 disclosed a high-severity Linux kernel vulnerability, dubbed 'Copy Fail' (CVE-2026-31431), that allows local users to gain root privileges. The bug arises from a...
OpenAI: ChatGPT Workspace Connector Write Actions Hit by Partial Disruption
On April 29, 2026, OpenAI identified an issue that automatically disabled write actions for some ChatGPT workspace connectors. The incident was classified as degraded performance, affecting the ChatGP...

Samsung Bespoke AI fridges will soon recognize thousands of food items via Gemini
Samsung is rolling out a software update this month to its 2024 and newer Bespoke AI fridges equipped with AI Vision cameras, upgrading object recognition from 37 common items to thousands. The jump c...
LlamaIndex CEO Jerry Liu claims the AI framework era is over, arguing that agent loops are now capab...
LlamaIndex CEO Jerry Liu claims the AI framework era is over, arguing that agent loops are now capable enough that context quality becomes the key competitive advantage, shifting focus from scaffoldin...

OpenAI model outperforms doctors in ER diagnoses, achieving 67% accuracy vs 55%
A peer-reviewed study compared an unnamed OpenAI model against physicians in simulated emergency room conditions, providing medical profiles and patient histories. The AI achieved a correct or close-t...
Alibaba's Metis agent slashes redundant AI tool calls from 98% to 2%, boosting accuracy
Alibaba introduced Metis, an AI agent trained with its new HDPO framework, which learns to skip unnecessary tool calls. The approach reduced redundant invocations from 98% to 2% while actually improvi...
OpenAI introduces Advanced Account Security with phishing-resistant login and stronger recovery options.
The update targets a growing vulnerability surface as enterprise adoption of ChatGPT and the OpenAI API accelerates. Phishing-resistant authentication—likely leveraging WebAuthn or hardware security k...
Citi launches internal agentic AI platform for employees
Citi is rolling out an internal AI platform that enables employees to create and deploy AI agents using top foundation models within a secure, bank-controlled environment. The move marks the bank’s fo...

FortyTwoMaru pitches agentic AI strategy for Chungcheong region digital transformation
Agentic AI startup FortyTwoMaru (42Maru) CEO Kim Dong-hwan presented a regional AX (AI transformation) strategy at the 2026 Daeil Forum in Daejeon, South Korea, targeting manufacturing, construction,...
ByteDance, Alibaba, and Tencent rush to acquire Huawei Ascend 950 AI chips after DeepSeek's launch
Chinese tech giants ByteDance, Alibaba, and Tencent are urgently acquiring Huawei's Ascend 950 AI chips, following the launch of DeepSeek's new AI model. The rush highlights growing demand for domesti...