AI News
Daily AI industry news — funding, products, policy, and major moves from global AI companies and startups.
Apertus: The Swiss Open Model Built for EU AI Act Compliance
The Swiss AI Initiative released Apertus as a fully transparent foundation model. It features 8B and 70B parameter versions supporting over
Why Anthropic is Now Requiring Government IDs for Claude
Anthropic now requires government IDs for certain Claude users. The verification process uses Persona to prevent platform abuse. This shift
Why NVIDIA Integrated OpenBao Into Its Serverless GPU Pipeline
NVIDIA officially adopted OpenBao for secret management in its cloud functions. The tool emerged as an open-source alternative to HashiCorp
Anthropic Fable 5 and Mythos 5 Hit by Trump Administration Export Controls
The Trump administration ordered Anthropic to disable Fable 5 and Mythos 5. Amazon researchers triggered the move by bypassing model guardra
Why Apple Intelligence in iOS 27 Moves Beyond the Chatbot
Apple integrates intelligence directly into core apps for iOS 27. The update shifts AI from conversational interfaces to embedded tools. On-
Gemma4-12B v2 Brings Local Coding Agents to 4.5GB VRAM
Gemma4-12B v2 enables local coding agents on 4.5GB of VRAM. The model improves tool-use benchmarks from 15% to 55% via CoT data. Quantized v
opencodex Unlocks Codex from OpenAI API Constraints
The opencodex local proxy enables Codex to connect with over 40 LLM providers. An SQLite migrator preserves session history during the trans
Bayer's PRINCE Platform Turns Decades of PDF Reports Into AI Agents
Bayer developed the PRINCE platform to automate preclinical data extraction. The system combines Agentic RAG and Text-to-SQL to process unst
Why Nobel Laureate John Jumper is Moving from DeepMind to Anthropic
Nobel laureate John Jumper is leaving Google DeepMind for Anthropic. Jumper led the AlphaFold project that revolutionized protein prediction
Why AI Coding Agents Are Creating a New Review Bottleneck
AI coding agents have shifted the development bottleneck to the review stage. Passing CI tests does not guarantee scalable or maintainable a
Why LLM-Generated AI Incident Reports Are a Dangerous Shortcut
LLMs risk erasing the critical thinking required for incident reports. Unlike code, narrative reports lack a deterministic test for accuracy
AI Deskilling: Why Medical and Coding Experts Are Losing Their Edge
AI dependency is causing professional deskilling in medicine and coding. An Anthropic study shows AI-assisted engineers score lower on conce
GLM-5.2 Opens 1 Million Token Context Window Under MIT License
GLM-5.2 introduces a 1 million token context window under an MIT license. The IndexShare architecture reduces per-token FLOPs by 2.9 times.
Why RTK's 90% Token Savings Might Be a Trap for Coding Agents
RTK claims to reduce coding agent token costs by up to 90 percent. The tool relies on fragile terminal parsing that risks silent failures. T
Alibaba's open-code-review Cuts Token Costs by 9x Over Claude Code
Alibaba open-sourced open-code-review for high-precision AI audits. The tool reduces token consumption to one-ninth of Claude Code. A hybrid
Cloudflare Temporary Accounts Remove the OAuth Barrier for AI Agents
Cloudflare introduced Temporary Accounts to automate AI agent deployments. The wrangler deploy --temporary command bypasses traditional OAut
How In the Weights Quantifies Digital Presence via AI Weights
In the Weights measures fame using internal AI model weights. The tool analyzes responses from models like GPT and Claude. Digital presence
KrIGF 2026 to Address HWP Compatibility and Font Copyright Hurdles
KrIGF 2026 will discuss legal solutions for HWP document compatibility. Polaris MCFG solves layout breaks by matching font metrics, not glyp
Why GPT-5.5 Hits an 86% Hallucination Rate
GPT-5.5 shows an 86% hallucination rate in new benchmarks. Open-weight GLM-5.2 outperforms larger models in reasoning efficiency. AI intelli
Norway Bans Generative AI for Students Aged 6 to 13
Norway is banning generative AI for students aged 6 to 13. The policy aims to protect basic literacy and numeracy skills. The government is
ASML Denies US Claims of EUV Equipment Leaks to China
The US government accuses ASML of leaking EUV equipment to China. ASML CEO Christophe Fouquet denies the claims and cites strict tracking. U
Why the US Government Banned Anthropic's Fable and Mythos Models
The US government banned Anthropic's Fable and Mythos AI models. National security concerns over jailbreaks triggered the export ban. Compan
The 4-Cent Email: How LLMs Automated Targeted Social Engineering
LLM-powered spear-phishing has reduced attack costs to 4 cents per email. Attackers now use session cookie theft to bypass traditional 2FA m
Claude Code Deciphers the Century-Old Mystery of Linear A
Tom Di Mino used Claude Code to decipher the ancient Linear A script. The researcher identified Linear A as an extinct Semitic language. AI
Smartbird: Allbirds' Radical Pivot to Data-Sovereign AI Infrastructure
Allbirds sold its footwear business for $43 million to rebrand as Smartbird. The company is now providing managed AI infrastructure focused
LangGraph and LangChain-core Vulnerabilities Expose AI Agents to RCE
LangGraph, Langflow, and LangChain-core face critical RCE and path traversal flaws. These vulnerabilities allow attackers to execute arbitra
Amazon Drops Sam Altman Biopic to Protect OpenAI Partnership
Amazon MGM canceled the distribution of the Sam Altman biopic Artificial. The move aligns with a $50 billion investment into OpenAI. Positiv
Why Reliance Built Jio Call Agent Directly Into the Network
Reliance Industries launched Jio Call Agent for 500 million users. The AI is embedded in the network to ensure AI sovereignty. A 110 billion
Hypernetworks: Generating Task-Specific Models 10-30x Cheaper in Real-Time
Hypernetworks generate task-specific model weights in real-time. Small specialized models are 10-30 times cheaper than general models. Nace
Match Group Survey Reveals the Hard Boundary for AI in Dating
Match Group found that 47 percent of US singles reject AI in romantic contexts. Users accept AI as a profile tool but reject it as an emotio