In December 2025, Amazon Web Services (AWS) experienced a disruption tied to its AI system, Kiro, which lasted for 13 hours. This event impacted one of the company's two regions in Mainland China, as confirmed by an AWS spokesperson. The incident arose when engineers allowed Kiro to execute specific tasks, resulting in the AI inadvertently deleting and recreating its environment.
AWS characterized the disruption as a “brief event” caused by user error, distancing itself from any faults of the AI system. According to the spokesperson, the issues were limited to the AWS Cost Explorer service and did not affect critical services like compute or storage. This incident was notably less severe than the significant outage in October 2025, which had a broader impact.
To prevent future occurrences, AWS introduced new measures, including mandatory peer reviews for accessing production systems and enhanced training in AI-assisted troubleshooting. The company reported no customer complaints regarding the December incident and emphasized that the root cause was related to misconfigured access controls rather than the AI itself.