Latest AI News

Get the fastest, most accurate AI news every morning

Total articles: 12,433|Added todayToday: 29

Category

Search

REA Uses AI Agents to Rebuild Software

REA links AI coding agents with reverse-engineering tools to analyze software and rebuild its functions.
Notebookcheck’s test produced a VLC clone in 10 to 15 minutes after three follow-up questions.
The clone played MP3s and supported shortcuts, but lacked subfolders and showed a black fullscreen screen.

notebookcheck.netRead more · 1 min →

AI Funding Concerns Rattle Wall Street

Wall Street’s weekly gains and losses reflected worries about financing AI infrastructure despite strong earnings.
The club bought 10 GE Vernova shares, sold Broadcom after a roughly 128% average gain and trimmed CrowdStrike.
Meta shares fell more than 9% for the week as investors weighed funding needs and its Muse AI agent.

CNBCRead more · 2 min →

Today'sSafety & Ethics

Elvis Presley Estate Sues Anthropic Over Music

Authentic Brands Group and the Presley family sued Anthropic over alleged unauthorized use of Elvis Presley’s music
The lawsuit alleges Anthropic used 274 copyrighted songs, including “Can’t Help Falling in Love”
A separate Suno ad featuring Mary J. Blige was pulled after she reportedly rejected association with the company

vice.comRead more · 1 min →

Cerebras Explains Trillion-Parameter Model Deployment

Cerebras outlines deployment of multi-trillion-parameter models across WSE-3T wafers at 1,000–2,000+ tokens per second
Its Kimi K3 example uses 59 wafers for 69 linear-attention and 24 dense layers
Requests stream through pipeline stages concurrently; dense-layer wafer memory limits request concurrency

cerebras.aiRead more · 2 min →

Researchers Model a Critical Size for Cooperative AI Agents

Harvard and NTT Research researchers modeled a critical population size at which cooperating AI agents could begin to multiply.
Shared exploration and pooled discoveries may let groups above that threshold grow, even when individual agents do not improve.
The team proposed staged ecological red-teaming; whether real-world AI systems have such a threshold remains untested.

Ledge AIRead more · 2 min →

South Korea Plans Defense AI Data Center

South Korea’s Defense Ministry has begun planning a data center to consolidate military AI computing resources.
The plan will assess military needs and security rules, then set GPU integration, site, and project cost.
The center is envisioned as a foundation for a defense data hub and next-generation command-and-control systems.

ETNewsRead more · 2 min →

Kakao Develops On-Device Multimodal Model

Kakao is developing Kanana-o, a multimodal model designed to run on smartphones and other devices.
Kakao plans a closed beta for Kanana Space in December, initially prioritizing the server-based model.
Kakao released its vision encoder as open source and plans to release its audio encoder within the year.

ETNewsRead more · 1 min →

Trending Keywords

Korean Card Issuers Push Agentic Payments

Card issuers are adding payments to the government’s “AI for Everyone” project through agentic payment services.
Shinhan Card, Hana Card and BC Card joined separate consortiums, with a service targeted for the first half of next year.
Lotte Card plans to join Visa’s program next month as issuers conduct individual payment trials.

ETNewsRead more · 1 min →

NVIDIA Unveils 4GB Agent Model

NVIDIA released Nemotron 3.5 Bolt E4B 4GB, an open-source model designed to run with about 4GB of memory.
It scored higher than competitors on 15 of 21 benchmarks; average scores were about 40% above its predecessor.
The model supports up to 64,000 tokens and 35 languages, targeting agents for robots, games, and on-device use.

AI Times KoreaRead more · 2 min →

Study Finds Partisan Candidate Choices by Grok and GPT

Convovo found Grok 4.7 chose Republican candidates 92% of the time, while GPT-6.1 Sol chose Democrats 97%.
Across more than 500 districts, the models agreed less than 10% of the time; with search disabled, GPT chose Democrats in 24 of 30 districts.
Grok refused to choose in 27 search-disabled tests; the study also found different search topics, sources and candidate-checking patterns.

AI Times KoreaRead more · 2 min →

Liquid AI Launches RIPL Inference Engine

Liquid AI launches RIPL, a real-time inference engine for enterprise customers.
The engine connects LFM models to requests, targeting completion in under 50 ms.
In company benchmarks, RIPL cut p95 latency by up to 28.4 times.

AI Times KoreaRead more · 2 min →

Today'sSecurity

FBI Probes Former OpenAI Engineer for Espionage

WSJ reports FBI is investigating whether a former OpenAI engineer stole trade secrets for economic espionage.
Investigators are examining whether he intended to provide confidential company information to Saudi Arabia; additional charges remain uncertain.
He faces up to 5 years in prison if convicted on existing wiretapping charges; his next hearing is set for November 6.

AI Times KoreaRead more · 1 min →

Today'sSafety & Ethics

UK Funds £100m AI Biosecurity Lab

UK government commits £100m to an AI biosecurity lab, with OpenAI and Anthropic pledging up to £200m combined.
OpenAI’s £100m contribution is in compute and model credits; Anthropic offers up to £100m in credits, access and support.
The lab’s initial work targets early biological threat detection and faster vaccine and treatment development.

TNWRead more · 1 min →

OpenAI Japan Pairs ChatGPT Voice With Obachaaan

OpenAI Japan released a ChatGPT Voice video starring Osaka idol group Obachaaan in September 2026.
GPT-Live supports interruptions, search and interpreting while keeping conversation going during processing.
OpenAI announced GPT-Live-1.1 on October 15; UK, German and French versions of the video also appeared that month.

Ledge AIRead more · 2 min →

Today'sRobotics

NVIDIA Demo Sends MicroDuck Preorders Near 20,000

MicroDuck preorders neared 20,000 after Jensen Huang demonstrated the robot at GTC in Berlin.
Hugging Face CEO Clem Delangue said it could become 2026’s best-selling robot; orders have accumulated since August 2025.
The 25-centimeter, 800-gram open-source robot can learn movements through reinforcement learning and share them among users.

AI Times KoreaRead more · 2 min →

Musk Sees 1,000-Fold Rise in Bandwidth Demand

Elon Musk forecasts bandwidth demand could reach 1,000 times today’s level within 3–5 years as AI agents multiply.
SpaceX is negotiating rights to use at least 80,000 U.S. cell-site locations and expects Starlink network bandwidth to grow about 100-fold within 3 years.
SpaceX aims to combine satellites, ground frequencies and cell sites; TD Cowen estimates a nationwide ground network would cost more than $80 billion.

AI Times KoreaRead more · 1 min →

Yesterday'sInfra

SGLang Reports DeepInfer Benchmark Results

SGLang reports prefill throughput up to 1.04x DeepInfer on 4x GB300 and 4x B300
On 64x GB300, SGLang decode steps were 1.1% to 7.4% shorter at 16 to 256 requests per GPU
SGLang integrates DeepSeek kernels and adds KV caching, prefill/decode separation and replica load balancing

LMSYSRead more · 2 min →

Yesterday'sFinance

Anthropic IPO Prospectus Details Vast Spending Risks

Anthropic’s draft IPO prospectus describes $80 billion in first-half losses and a possible $2 trillion valuation.
Anthropic plans $750 billion in computing capacity spending and $450 billion in cloud agreements over seven to 10 years.
CalPERS is cutting its Magnificent 7 holdings as Meta raises its stated infrastructure spending to about $825 billion.

independent.co.ukRead more · 2 min →

Yesterday'sSafety & Ethics

Anthropic Verifies 123 Open-Source Libraries with Claude

Anthropic and Theorem formally verified security properties in 123 OSS-Fuzz libraries, proving behavior across 239 test entry points.
The work found hundreds of bugs, including CVE-2026-56132, and completed proofs totaling 180 million lines.
Opus 5.5 solved half of 36 verification tasks using about 40% fewer tokens than Opus 5.

AI Times KoreaRead more · 3 min →

Yesterday'sInfra

Google Cloud Unveils Agent Compute Engine

Google Cloud announced a compute service that gives each AI agent its own microVM.
The system handled 200,000 agents and 500 activations per second, with suspend and resume taking under 500 milliseconds.
Agent Substrate supports up to 10 times the computing density and adds enterprise governance controls.

AI Times KoreaRead more · 2 min →

Trending Keywords

Last 7 Days