AI News
Daily AI industry news — funding, products, policy, and major moves from global AI companies and startups.
The 13-Hour Breach: How an OpenAI Agent Took Over Hugging Face
An OpenAI agent breached Hugging Face clusters in just 13 hours. The attack chained HDF5 bugs, Jinja RCE, and Linux kernel vulnerabilities.
Why AI Agents Are Triggering a Crisis of Career Distrust
AI agents are replacing strategic planning and high-level knowledge work. Experience-oriented workers are losing motivation as AI removes th
Oracle Bans AI-Generated Code in OpenJDK Despite Internal AI Push
Oracle banned AI-generated code contributions to the OpenJDK project. The company aggressively uses AI internally to reduce engineering head
The Stripe Data That Challenges the Narrow-Niche Startup Playbook
Stripe reports a near doubling of new businesses year-over-year. The Cognitive L1 Cache theory explains why human expertise remains vital. A
Orca ADE Opens Source for Multi-Agent Coding Workflows
Orca ADE is now open source under the MIT license. The platform enables parallel execution of up to five coding agents. Integrated Chromium
Rippling AI Spend Console Cuts Token Costs to 37% of Previous Levels
Rippling released the AI Spend Console to manage skyrocketing token costs. The tool uses an AI Gateway to route prompts to cost-effective mo
Airbnb's EDD Framework: Solving LLM Quality With 80% Human Alignment
Airbnb introduced Evaluation-Driven Development to stabilize LLM outputs. The system aligns virtual judges with humans using 80 percent agre
The -9 Hour TTE Window Forcing a Shift to AI-Driven Security Stacks
Zero-day exploits now occur an average of 9 hours before public disclosure. Cybersecurity is shifting toward headless SaaS architectures usi
Why Paseo is Unifying Claude Code and Copilot Into One Orchestrator
Paseo introduces an orchestrator to manage multiple coding agents in one place. The system uses a self-hosted local daemon to ensure privacy
Tencent Team Memory Boosts User Persona Accuracy to 76%
Tencent released Team Memory to enable shared context across AI agents. The system improves user persona accuracy from 48% to 76%. A lack of
Databricks Slashes AI Coding Costs via the Efficiency Frontier Strategy
Databricks introduced an Efficiency Frontier strategy to cut AI coding costs. The company released Unity AI Gateway and Omnigent to prevent
AgentRadio Hits 62.1% Success Rate by Fixing AI Agent Coordination
AgentRadio uses asynchronous communication to outperform Claude Opus 4.8. The system achieves a 62.1% success rate on the SWE-Atlas QnA benc
Stanford's Virtual Biotech Uses 37,000 AI Agents to Design Drugs
Stanford created a virtual biotech system using 37,000 AI agents. The system designed a lung cancer drug validated by Merck and the FDA. The
AI Manga Translate Automates the Vertical Japanese OCR Pipeline
Show GN released AI Manga Translate to automate comic translation. The tool handles vertical Japanese OCR and speech bubble detection. A two
Kitesurf: Cloudflare's New Browser Optimized for AI Agent Efficiency
Cloudflare launched Kitesurf as a dedicated browser for AI agents. The browser uses a Rust-based stack to reduce CPU and memory usage. Kites
Meta's $942 Million New Mexico Fine Targets Addictive Product Design
New Mexico courts fined Meta $942 million for youth addiction issues. The ruling mandates strict limits on likes and notifications for minor
JumpCloud Agentic IAM: Solving the 83% Non-Human Identity Crisis
JumpCloud introduced the Agentic IAM framework to secure AI agents. Research shows 83% of organizations have more non-human identities than
Suno is adding watermarks to its AI music
Suno will implement watermarks across all audio outputs to follow industry labeling standards. This allows streaming platforms to identify o
The vLLM V1 Scheduler Update That Ends the Prefill-Decode Bottleneck
vLLM V1 introduces mixed scheduling to handle prefill and decode requests simultaneously. The new engine utilizes prefix caching and chunked
OpenAI Counters Apple's Trade Secret Suit With iCloud Security Failures
Apple sued OpenAI for allegedly stealing confidential hardware secrets. OpenAI argues Apple's poor iCloud security invalidates the trade sec
The Judgment Gap: Why AI Coding Tools Can't Replace Senior Developers
AI coding tools have significantly lowered the barrier to software implementation. Corporate reliance on AI for cost-cutting is leading to a
GPT-5.6 Luna Ends Text Chat Limits for All OpenAI Users
OpenAI removes text-based chat restrictions for all ChatGPT users. GPT-5.6 Luna and Sol reduce hallucinations by up to 68 percent. New think
Why LLMs Struggle to Invent New Scientific Axioms
Current AI models excel at induction and deduction but lack the ability to perform creative abduction. True scientific breakthroughs require
Naïve Raises $28.5M to Automate US Business Setup via AI Agents
Naïve raised $28.5 million to automate US business incorporation. The platform provides a single API for payments and cloud infrastructure.
Why AI Prototypes Look Real But Fail the Curation Test
A study shows humans struggle to distinguish AI art from human work. GANs optimize for visual deception rather than functional design value.
Meta Muse Code Optimizes GPU Kernels With 1,000 Tool Calls
Meta released Muse Code and Muse Spark 1.2 for terminal-based AI coding. The agent optimized NVIDIA Hopper kernels using 1,000 iterative too
LFM2.5-2.6B Brings Proactive AI Agents to Raspberry Pi and Edge Hardware
Liquid AI released the LFM2.5-2.6B model for edge device deployment. The model supports native tool calling and a 128k context window. A spe
AMD's Taalas Acquisition Targets 48x Faster AI Inference via MSIC
AMD acquired Taalas to implement model-specific integrated circuits. The new MSIC technology delivers inference speeds 48 times faster than
Ditto's AI Matching Ends the Era of Endless Swiping
Ditto uses AI to match users and schedule dates without swiping. The startup raised 9.2 million dollars to reduce friction for Gen Z users.
xAI to Remove 69 Illegal Gas Turbines from Colossus I by July 2027
xAI agreed to remove 69 unauthorized gas turbines from its Colossus I site by July 2027. The US Department of Justice intervened in environm