#aisecurity

⚠️ Inside the Metabase SQLi: Exploited in the Wild https://www.wiz.io/blog/inside-the-metabase-sqli-exploited-in-the-wild (wiz.io)

🔥 INCIDENT REPORTING OpenAI’s Upcoming Astra Model Raises Autonomous Cyberattack Concerns https://www.securityweek.com/openais-upcoming-astra-model-raises-autonomous-cyberattack-concerns/ (securityweek.com)

🔥 INCIDENT REPORTING Three interviews: system fragility, operational clarity, and Identity for AI agents - ESW #471 https://www.youtube.com/watch?v=fDr0Q8Lc8dk (youtube.com)

Skill or Shortcut? You Can Now Get the Right Answer Without Building the Skill | CyLab Security Academy (cylabacademy.org)

What does a competition score prove when AI can help solve a hard cybersecurity challenge in minutes? Our recent white paper examines six years of competition data and the growing gap between performance and expertise. cylabacademy.org/posts/2026-0... #Cybersecurity #AISecurity @cmu.edu

AI Agent Code Isn't a "Trusted Product" Anymore. Kubernetes Sandbox Design Has an Answer (kubo.hexabase.io)

AI agents now write AND run code you never reviewed. Treating that as a "trusted product" is a security gap—and plain container isolation won't save you. Here's why Kata Containers' microVM boundary is becoming table stakes for agent sandboxes.