Five days, four platforms: correlated failures and AI agent concentration risk
Five Days, Four Platforms: Correlated Outages and AI Agent Concentration
The Cloud Security Alliance (CSA) published a research note analyzing the impact of correlated failures at Microsoft 365, Anthropic, OpenAI, and AWS from November 12 to 15, 2026, on AI agent deployments. A nearly 24-hour gateway failure at Microsoft 365 simultaneously caused errors in Copilot and in the Microsoft 365 connectors of Anthropic and OpenAI, and multi-model redundancy failed at the shared data layer; Anthropic had a login incident and then a broader failure on November 13, the latter causing scheduled tasks to be delayed by more than two hours and some to be skipped; the AWS us-east-1 incident lasted about three hours and affected IAM Identity Center logins and the AI development environment Kiro.