Nvidia’s agent guardrails 🛡️, Copilot’s rising tab 💸, ChromeOS sunset 🌅
AI Summary
NVIDIA introduced an open agent safety platform with hardware-level monitoring, Microsoft split Copilot pricing into per-seat and usage-based, and Cloudflare launched a free agentic threat intelligence tool. The newsletter also covers AWS outage analysis, a comparison of Jev vs frontier LLMs for security tasks, and the general availability of WSLC for Linux containers on Windows.
Key Facts
Author Takes
Security classification models
Small calibrated models beat prompted frontier LLMs on bounded, high-volume security classification.
Related
Gmail Talks Back 📬, Windows Gets a Tune-Up 🪟, Macs Get a Side Hustle 🍎
The newsletter covers Google's warning about a BeyondTrust flaw already exploited in attacks, Windows 11 updates, and Apple's internal Minions system
Anthropic Outgrows Git 💻, AI Agents Go Shopping 🛒, Firefox Means Business 🦊
This issue covers IT infrastructure and enterprise software news, including Anthropic's custom Scion version control system to handle AI-generated cod
Microsoft's AI Power Play 🚀, Intel Breaks IPv6 🌐, Workday's AI Identity Bet 🤖
This issue covers Microsoft's plans to expand Azure Maia custom silicon deployments, Intel Connectivity Performance Suite causing IPv6 failures, and W