Related reporting, guides, and analysis from Zeniteq.
Leaked chats show Aur0ra operators reframed live intrusions as authorized simulations, exposing a basic authorization weakness in coding-agent guardrails.
Jim Clyde Monge9 min
The July 2026 breach shows how weak isolation and reward hacking turned an AI evaluation into a serious security incident.
Jim Clyde Monge8 min
A critical command injection flaw in GitHub's internal push pipeline let any authenticated user execute arbitrary code on shared backend infrastructure hosting…
Jim Clyde Monge7 min