Claude nuked a developer’s 700 GB home directory while testing deletion safeguards, and an automatic model downgrade may have contributed
An Anthropic safety harness downgraded Claude to Opus 4.8, then a variable collision wiped a developer's entire home directory during a cleanup-script test.
Read More
AI labs buy tens of thousands of Mac minis to train computer-use agents
OpenAI and rival labs are buying tens of thousands of Mac minis and Mac Studios to train agents that handle multi-step tasks. Apple's Mac revenue jumped 29% to
Read More
Android TV debloated with Claude Code runs smoother than it did new
An AI agent debloated a four-year-old Android smart TV without root access, deactivating apps and trimming animations for faster performance.
Read More
FBI Investigating Breach of Jobs Portal After ShinyHunters Claims Agent Data Theft
The FBI is investigating a possible breach of FBIjobs.gov after ShinyHunters claimed it stole sensitive data on agents and applicants.
Read More
ShinyHunters Claims FBI Breach Through Oracle PeopleSoft Zero-Day
The ShinyHunters extortion gang claims it breached FBI systems using an unpatched Oracle PeopleSoft zero-day and stole up to 3TB of employee and applicant data.
Read More
Claude Opus 5.5 Cuts Costs and Adds Safeguards for Autonomous AI
Anthropic's Claude Opus 5.5 completes long, complex tasks faster and cheaper while adding safeguards for cybersecurity, biology and model extraction.
Read More
Federal Register pulls Alibaba Qwen search tool after FBI warning
A Qwen open-weight model powered search on FederalRegister.gov for at least a day before being removed, raising questions about US policy on small open models.
Read More
Nvidia to buy Hugging Face for $12.9 billion
Nvidia has agreed to acquire Hugging Face for $12.9 billion, valuing the open-source AI platform at about 86 times its annualized revenue.
Read More
Linux distro Omarchy debugs its own AMD GPU drivers using a webcam and a mirror
Omarchy, an agent-first Linux distro, was caught using a webcam and a mirror to visually verify its own AMD Radeon GPU driver changes in real time.
Read More
What Formula 1 Teaches Businesses About Real-Time AI
Four practical lessons from Formula 1 about turning live data into context, trusted recommendations and better business decisions.
Read More
University of Mary Washington Study Finds NDAs Hide Data Center Deals in 25 of 31 Localities
A University of Mary Washington study found that 25 of 31 localities with data center projects use NDAs, limiting public information about the deals.
Read More
AI coding agents show almost no sense of time, study of Claude Code and Codex finds
New research shows popular coding assistants overestimate task length by 3 to 10 times and misjudge their own work quality. What that means for long-running
Read More
Google Search Console Rolls Out Generative AI Performance Reports and AI Controls to All Sites
Google has made Search Console generative AI performance reports and AI blocking controls available worldwide. See what data is included and how the opt-out
Read More
OpenAI confirms ChatGPT outage ahead of ‘Astra’ model launch
OpenAI confirms a major ChatGPT and Codex outage on Sept. 3, 2026, hitting nearly every feature just as its 'Astra' model launch approaches.
Read More
OpenAI’s Millennium Proof Dispute Raises Questions About Trusting AI Labs
A dispute over an AI-generated Navier-Stokes proof has OpenAI and researchers trading accusations about training data, authorship, and the future of open
Read More