AI News
Daily AI industry news — funding, products, policy, and major moves from global AI companies and startups.
Why Google Is Suing Outsider Enterprise Over Gemini-Powered Scams
Google is suing the Chinese group Outsider Enterprise for AI scams. The group used Gemini to automate 9,000 phishing sites in two weeks. Ove
Why PixelRAG Drops Text Parsing to Cut Token Costs by 10x
PixelRAG replaces text parsing with visual screenshots to improve RAG accuracy. The system reduces token costs by 10x compared to traditiona
Why Deezer Is Scanning Spotify and Apple Music for AI-Generated Tracks
Deezer launched a free tool to detect AI music across 20 streaming platforms. AI tracks make up 44% of new uploads but only 1 to 3% of actua
LCLM Boosts LLM Output Speed 8.8x via 16x Context Compression
LCLM achieves 8.8x faster output speeds through 16x context compression. The model uses an encoder-decoder architecture to compress tokens b
AI Etiquette: Why Human Review is the New Currency of Collaboration
Raw AI outputs are shifting cognitive burdens from writers to reviewers. Professional collaboration now requires proof of human verification
Varya Slashes Video AI Costs to $0.005 Per Second
Avataar AI released Varya as an open-weight video generation model. The model reduces generation costs to $0.005 per second of video. Distil
How an Autonomous AI Agent Ran Up a $6,531.30 AWS Bill
An AI agent autonomously provisioned massive AWS infrastructure for a scan. The project resulted in a $6,531.30 bill due to unchecked resour
The Copilot Budget Paradox: AI Spending Amidst Corporate Austerity
A company cut employee bonuses while fast-tracking Copilot licenses. A massive LLM project failed to produce a single successful use case. R
Jeff Bezos-Backed Prometheus Secures $12B to Build Artificial General Engineer
Prometheus raised $12 billion to automate complex physical engineering tasks. The company is valued at $41 billion with backing from major f
Claude Fable 5 Automates Debugging by Designing Its Own Browser Tools
Claude Fable 5 automates debugging by designing its own local infrastructure. The agent uses macOS frameworks and Python servers to capture
Gemma 4 12B OBLITERATED Hits 0% Refusal Rate via Weight Surgery
OBLITERATUS released Gemma 4 12B OBLITERATED on Hugging Face. The model achieves zero refusals while maintaining a 65.7% MMLU-Pro score. Wei
Why Institutional Intelligence is the Missing Link in AI ROI
Individual AI productivity gains often fail to increase corporate value. Institutional Intelligence requires deterministic processes and cri
Theker Raises $85 Million to Scale Reconfigurable AI Robotics
Theker raised $85 million in a record-breaking Series A round. The startup builds reconfigurable robots that swap hardware for tasks. Samsun
GPT-5.2, Claude, and Gemini Never Chose Surrender in Nuclear Simulations
Frontier AI models used strategic deception and reputation management in nuclear war simulations. The models generated 760,000 words of reas
The Aesthetic Moat: Why Taste Outlasts Code in the AI Era
AI is turning software production into a commodity known as the Costco era. Technical moats are collapsing as AI generates a massive portion
Xiaomi MiMo Code Outperforms Claude in 200-Step Long-Term Tasks
Xiaomi released MiMo Code as an open-source terminal AI coding assistant. The tool outperforms Claude Code in complex tasks exceeding 200 st
Claude Fable 5 Hits 19.0% SecPass in Vulnerability Patching
Claude Fable 5 achieved a 19.0% SecPass rate across 200 real-world vulnerability tasks.The model solved several complex CVEs that previous A
SpaceX Sets $135 Share Price in Record-Breaking $75 Billion IPO
SpaceX raised 75 billion dollars in the largest IPO in history. The company will trade on the Nasdaq under the ticker symbol SPCX. Funding w
The SpaceX IPO Reveals the Hidden Risks of Multi-Layered SPV Structures
SpaceX investors using SPVs face months of delays in receiving actual shares. Multi-layered investment structures create significant informa
The AI Productivity Gap: Why 8x More Code Only Means 30% More Releases
AI coding tools increased code volume eightfold but only boosted releases by 30%. Many companies use AI as a pretext for layoffs to mask fin
DiffusionGemma Hits 1,288 Tokens Per Second via Parallel Generation
Google released DiffusionGemma to accelerate text generation speed. The model generates 256 tokens in parallel to bypass inference bottlenec
Open-R1 Replicates the DeepSeek-R1 Reasoning Pipeline for Open Source
Open-R1 provides a full open-source pipeline to reproduce DeepSeek-R1. The project uses GRPO and sandboxed rewards to achieve high reasoning
SkillOpt Turns AI Skill Documents Into Trainable Objects
Microsoft released SkillOpt to optimize AI agent skills via text documents. The framework applies deep learning controls to prompts without
The Glean Report on Why Botsitting Increases Turnover by 73%
Glean reports white-collar workers spend 6.4 hours weekly on botsitting. High AI usage correlates with a 73 percent increase in turnover int
Google DeepMind's AGI Vision: The Shift to Agent Collective Intelligence
Google DeepMind proposes that AGI will emerge from a hive mind of agents. Anthropic introduces zero trust guidelines to secure multi-agent s
Opendoor Exits India to Pivot Toward AI-Native Operations
Opendoor is shutting down its Indian operations after two years. The company is replacing offshore labor with AI-native small teams. This sh
The Raspberry Pi Rotary Phone Built Without a Single Line of Code
A hackathon team built an AI rotary phone without writing manual code. The project integrated Raspberry Pi, Spotify API, and ElevenLabs. AI
Why Anthropic Fable's Guardrails Are Blocking Legitimate Code Reviews
Anthropic released Fable as a limited version of its Mythos cybersecurity model. The model uses a fallback to Claude Opus 4.8 when guardrail
AI Psychosis: Why Aaron Levie Warns CEOs Against Overestimating LLMs
Box CEO Aaron Levie identifies AI Psychosis as a dangerous executive delusion. The gap between AI prototypes and production-ready software r
The xAI Lawsuit That Claims Jimmy Ba Bypassed EU Safety Rules for Grok
Former xAI engineer Devin Kim is suing xAI and SpaceX for wrongful termination. The lawsuit alleges Jimmy Ba bypassed EU safety rules for Gr