OpenAI Silently Re-Enabled Codex Training: Fix It Now
OpenAI re-enabled the Codex model training toggle for opted-out users on November 3 without notification. There are two independent settings to disable. Here ...
Red Hat OpenShift 5: Lightspeed Agent, MCP Server, Post-Quantum Crypto
Red Hat OpenShift 5 ships at KubeCon 2026 with an autonomous Lightspeed agent, built-in MCP ...
Zarf 1.0: Air-Gapped Kubernetes Delivery Goes Stable
Zarf 1.0 ships at KubeCon 2026 with a new v1beta1 schema, keyless Sigstore signing, and ...
Google Gemini Exposes Gmail and Drive to Human Reviewers
Google reversed its Gemini privacy promise: Gmail and Drive data is now routed to human ...
Google ADK 2.14: A2A Stable, 3 Breaking Changes to Fix
Google ADK 2.14 graduates A2A and 42 modules to stable but ships 3 breaking changes ...
CrowdStrike AI Kill Switch Stops Rogue Agents Dead
CrowdStrike added an AI Kill Switch to Falcon Guardian. One action terminates a rogue agent, blocks relaunch, and cuts network access. Here's how ...
Debian and Ubuntu APT Flaw Lets MitM Attackers Swap Verified Packages and Run Code as Root — Patch Now
CVE-2026-109518 lets attackers swap a verified APT package after the signature check and install it ...
GitHub Actions Blocks pull_request_target — Fix by December 2
GitHub is blocking pull_request_target by default in public repos on December 2. The first brownout ...
Outlook Is Rejecting Your Emails Since Nov 5 — Fix It Now
Microsoft's Outlook.com silently rejects email from domains without DMARC since Nov 5. Here is the ...
OpenAI Hired the Hackers Who Breached It — Your API Is Next
OpenAI just hired the security firm that broke into it. On November 6, Hacktron — ...