OpenAI Launches GPT-6 Cyber: Its Most Powerful Hacking Model Yet — for Vetted Defenders Only
OpenAI has launched GPT-6 Cyber, a security model scoring 48.5% on ExploitGym and rated Critical — but only approved Daybreak Red teams can use it.
-
-
OpenAI’s AI Built 948 Working Linux Root Exploits in One Run — for $1,279 Each
OpenAI’s autonomous pipeline found 948 Linux kernel flaws giving ordinary users root access, for $1,279 each. Here’s what it means for patching.
-
CrowdStrike’s AI Kill Switch: How It Stops Rogue Agents — and the $495 Million Deals Behind It
CrowdStrike’s AI Kill Switch lets admins stop rogue AI agents, and its reported $495M Fiddler and Dreadnode deals add monitoring and pre-launch testing.
-
Chainalysis Breach: AI Agents Used to Break Into the Firm That Tracks Crypto Crime
Chainalysis says an attacker using ‘offensive agentic techniques’ accessed customer names and emails. What was taken, what wasn’t, and what to watch for.
-
PHP2Shell: Two AI-Discovered PHP Flaws Trigger ‘Patch or Unplug’ Warnings From the NSA and FBI
The NSA and FBI have told organizations to patch two critical PHP flaws dubbed PHP2Shell — or take systems offline. The AI-discovered bugs threaten much of the web.
-
70% of Companies Now Let AI Make Human Decisions — but Half Can’t Audit What It Does
70% of organizations now let AI make decisions that once needed human approval — yet only 49% can audit every AI action. Key findings from Illumio’s 2026 report.
-
OpenCode API Key Breach: What Happened, Who Is Affected, and How to Rotate Your Keys
Anomaly Innovations confirmed a breach of OpenCode Console exposing stored OpenAI, Anthropic, and Google API keys. Here’s what happened and how to rotate your keys.
