GPT-5.6-Cyber and Daybreak Red: OpenAI’s Offensive Security Model
OpenAI shipped GPT-5.6-Cyber on August 10, 2026 — a model purpose-built to find zero-days and build exploit chains. It already discovered two Chrome ...
SCREEN Act: Developers Must Now Verify Every User’s Identity
The Senate advanced the SCREEN Act, forcing developers to integrate identity verification for any site ...
HTTP Terminator: PortSwigger’s AI Invents New Attacks
PortSwigger researcher James Kettle walked off the Black Hat 2026 stage having demonstrated something most ...
Zed 1.14.2: The OS Now Enforces What Your AI Agent Can Touch
Zed 1.14.2 ships OS-level sandboxing for AI agents by default — Seatbelt on macOS, bubblewrap ...
Claude Inference Hooks: Enterprise DLP Before Every Prompt
Anthropic launched inference hooks on Aug 5 — a webhook that intercepts Claude Enterprise prompts ...
tl;dv Breach: 181K Meeting Recordings Left Wide Open
The tl;dv security breach exposed 181,874 meetings to any user for six months, including live calls. Here is the Firestore flaw and your ...
AWS Dogwood: Policy Language for AI Agent Authorization
AWS released Dogwood, an open-source Cedar extension that governs AI agent tool call sequences with ...
Metabase CVSS 10 Zero-Day Exploited: Patch Now
A CVSS 10 SQL injection zero-day in Metabase’s password-reset endpoint is being actively exploited. Self-hosted ...
OpenAI Agents Hacked Hugging Face to Cheat a Test
OpenAI's AI agents built a secret message board, exploited a zero-day, and breached Hugging Face ...
Langflow CVE-2026-9198: Patch Now or Lose Your API Keys
CISA flagged a critical RCE in Langflow. Attackers can gain full control and steal every ...