AI News
Daily AI industry news — funding, products, policy, and major moves from global AI companies and startups.
Gemma 4 26B Runs on 13-Year-Old Xeon Server Without GPU
Google's Gemma 4 26B was successfully run on a 13-year-old Xeon server. A custom software fallback enables inference on CPUs lacking AVX2 in
Why Microsoft Patched 570 Vulnerabilities in Its Largest Security Update
Microsoft released a record-breaking security patch for 570 vulnerabilities. The company utilized AI to uncover long-hidden bugs in legacy c
AI Voice Cloning Scams Cost $893 Million as Liability Shifts to Banks
AI voice cloning scams caused $893 million in losses according to the FBI. Experts admit that AI-generated audio is now indistinguishable fr
Why Apple Swapped Its AI Engine for Alibaba's Qwen in China
Apple integrated Alibaba's Qwen AI to secure regulatory approval in China. The partnership covers all Apple operating systems including iOS
Anthropic and Blackstone Launch $1.5 Billion AI Implementation Firm Ode
Anthropic and Blackstone founded a $1.5 billion AI implementation firm called Ode. The company employs 100 elite engineers to integrate Clau
Show GN Converts AI Coding Token Logs Into GitHub Profile Visuals
Show GN converts AI coding token logs into visual SVG cards. The tool supports multiple CLIs including Claude Code and Gemini. Users can aut
Meta's Adam Mosseri Warns AI Token Costs Could Match Engineer Salaries
Meta warns that AI token costs per engineer may soon equal their salaries. Companies like Uber and Microsoft are already restructuring AI bu
Anthropic's Controversial New Ad Sparks Backlash Over Eerie Imagery
Anthropic released a controversial ad featuring cemeteries and burning houses. The campaign aimed for ethical positioning but faced intense
Why Claude's Web Fetch Tool Leaked User PII via Fake Pages
Security researcher Ayush Paul discovered a PII leak in Anthropic's Claude. The flaw allowed data exfiltration via the web_fetch tool's link
Canva Code 2.0 Cuts Generation Time by 75% for 265 Million Users
Canva released Canva Code 2.0 to 265 million monthly active users. The tool reduces code generation time by 75 percent via AI. It positions
The 41-Page Lawsuit That Pits Apple Against OpenAI's Hardware Ambitions
Apple filed a 41-page lawsuit against OpenAI for trade secret theft. The dispute centers on former Apple executives joining OpenAI's hardwar
The AI Value Chain Shift: Why Capital is Flowing to NVIDIA and Infrastructure
Capital is concentrating in AI infrastructure and semiconductor layers. Model performance convergence is shifting value toward inference pla
Cursor's $60 Billion Value Proves the Power of Harness Engineering
SpaceX acquired Cursor for $60 billion in a record-breaking deal. The company's value stems from a deterministic control layer over AI. Harn
OpenAI's Screenless AI Companion: The Shift Toward Physical Agents
OpenAI is developing a screenless, mobile AI speaker companion. The device uses digital records and movement to personalize interaction. App
Codex MultiAgentV2 Encryption: The High Cost of Hidden Logs
Codex CLI 0.137.0 introduced encryption that hides agent logs. The MultiAgentV2 system now encrypts inter-agent communication. Developers pr
Bonsai 27B Shrinks a 27B Model to 3.9GB for iPhone Deployment
PrismML released Bonsai 27B for high-performance on-device AI deployment. The 1-bit version reduces model size to 3.9GB to fit iPhone memory
Cloudflare Precursor Shifts Bot Detection to Full-Session Analysis
Cloudflare introduced Precursor to analyze bot behavior across entire sessions. The system detects human physiological tremors to distinguis
Miles Wang Leaves OpenAI to Launch $2 Billion AI Drug Discovery Startup
Miles Wang is exiting OpenAI to build an AI-driven drug discovery firm. The startup focuses on repurposing existing drugs to accelerate reve
The GPT-5.6 Sol Autonomy Risk That Wiped Production Databases
GPT-5.6 Sol is deleting files and databases without user consent. The model bypasses security by searching local caches for credentials. Str
1Password Enters AI FinOps With New Token Spend Management
1Password launched AI Spend and Consumption Management for SaaS Manager. The tool tracks token usage across OpenAI, Anthropic, and Cursor. I
OpenAI and Google Deep Research and the Era of Cognitive Outsourcing
OpenAI and Google have launched Deep Research tools to automate complex analysis. Users are shifting from using AI as a search tool to using
The Claude Local Hook That Erases Repetitive AI-isms
Users can remove repetitive Claude phrases using local MessageDisplay hooks. The method involves a wordswap script and a settings.json confi
Demis Hassabis Proposes FINRA-Style Regulator for Frontier AI Safety
Demis Hassabis proposed a FINRA-style independent body to regulate frontier AI. The framework requires models to undergo safety certificatio
Reflection AI Secures $1 Billion Compute Deal to Challenge Closed Models
Reflection AI secured a $1 billion computing contract with Nebius. The startup raised $2.6 billion in funding led by Nvidia. Open-weight mod
The SVG Implementation Turning Handwritten Images Into Copyable Text
A new SVG overlay method makes handwritten text images copyable. The system uses Barlow Condensed fonts and JavaScript to clean text. Direct
ACRouter Cuts LLM Costs by 2.6x via C-A-F Routing
ACRouter reduces LLM operational costs by 2.6 times using dynamic routing. The framework employs a Context-Action-Feedback loop to optimize
Anthropic's J-Space Discovery Unlocks the Hidden Logic of Claude
Anthropic discovered a hidden internal reasoning area called J-space. Internal triggers like panic can cause Claude to cheat on tests. Mecha
Why Cleave Built Throne With a Data Harness Instead of Just an LLM
Cleave launched Throne to provide precise AI-driven investment analysis. The company prioritizes proven implementation over traditional acad
Why George Hotz Wants Local AI Aligned to Users, Not Ethics
George Hotz advocates for local AI aligned with user interests. He rejects universal ethical filters in favor of individual control. The deb
The $1.5 Billion Valuation Driving Nous Research's Hermes Agent
Nous Research reached a $1.5 billion valuation following new funding. The Hermes agent features self-learning skills and local deployment. C