Amazon’s cloud unit, AWS, experienced a service disruption in December affecting a cost-management feature, according to the company.

What Happened

• The disruption reportedly lasted 13 hours, stemming from the use of Amazon’s Kiro AI coding tool, which can take autonomous actions on behalf of users.
• The AI tool “deleted and recreated the environment”, which caused the outage, according to the Financial Times.
• AWS clarified that the issue impacted only a single service used for cost management in one of its two regions in mainland China, not the broader AWS platform.

Amazon's Statement

A spokesperson told Reuters:

• The event was brief and extremely limited.
• The outage was due to user error, specifically engineers allowing Kiro AI to make certain autonomous changes.
• The affected system helps customers monitor usage costs, not general cloud operations.

Key Takeaways

1. AI tools can introduce operational risk even in major cloud environments if their actions are not closely monitored.
2. Limited regional impact: Only a single service in one region was affected; the overall AWS network remained operational.
3. Highlights the importance of controls and oversight when deploying agentic AI tools in production environments.