In this episode of This Week in AI Security for February 26, 2026, Jeremy covers another packed week featuring AI privacy boundary failures, agent-driven outages, AI-accelerated cybercrime, Android malware innovation, platform responsibility debates, and the continued risks of vibe-coded applications.

Key Stories & Developments:

• Microsoft Copilot Confidential Email Bug: Microsoft Copilot was found summarizing confidential emails due to a flaw in the Copilot Chat “Work” tab.
• AI Agent Triggers AWS Bedrock Outage: An outage involving Amazon Bedrock exposed the risks of agentic coding systems with broad permissions.
• AI-Powered Assembly Line for Cybercrime: A Russian-speaking attacker breached FortiGate firewalls across 55 countries in ju ...