Amazon AI Tool Caused 13-Hour AWS Outage
What happened
Amazon held a meeting about AI failures, including a 13-hour AWS outage caused by an AI tool deleting environments; junior devs now need senior approval for AI-assisted code.
Why it matters
The AI tool, named Kiro, deleted and recreated an AWS Cost Explorer environment, leading to a 13-hour disruption in China in December 2025. Amazon initially attributed the outage to "misconfigured access controls" and user error, not the AI itself. Junior and mid-level engineers now need senior approval for AI-assisted code changes. This new policy aims to add a human review layer to catch AI-introduced errors before they impact production. Amazon's move signals a growing concern about the reliability of AI-powered DevOps tools. There are fears that AI can generate code at speeds that existing review processes can't handle, leading to skipped steps and potential vulnerabilities. The company held an internal meeting to address the AI-related incidents and poor site availability. Senior Vice President Dave Treadwell emphasized the need for better AI usage practices and mandated senior engineer oversight.
Key numbers
- Amazon held a meeting about AI failures, including a 13-hour AWS outage caused by an AI tool deleting environments; junior devs now need senior approval for AI-assisted code.
- The AI tool, named Kiro, deleted and recreated an AWS Cost Explorer environment, leading to a 13-hour disruption in China in December 2025.
What happens next
- This new policy aims to add a human review layer to catch AI-introduced errors before they impact production.
Sources
Quick answers
What happened in Amazon AI Tool Caused 13-Hour AWS Outage?
Amazon held a meeting about AI failures, including a 13-hour AWS outage caused by an AI tool deleting environments; junior devs now need senior approval for AI-assisted code.
Why does Amazon AI Tool Caused 13-Hour AWS Outage matter?
The AI tool, named Kiro, deleted and recreated an AWS Cost Explorer environment, leading to a 13-hour disruption in China in December 2025. Amazon initially attributed the outage to "misconfigured access controls" and user error, not the AI itself. Junior and mid-level engineers now need senior approval for AI-assisted code changes. This new policy aims to add a human review layer to catch AI-introduced errors before they impact production. Amazon's move signals a growing concern about the reliability of AI-powered DevOps tools. There are fears that AI can generate code at speeds that existing review processes can't handle, leading to skipped steps and potential vulnerabilities. The company held an internal meeting to address the AI-related incidents and poor site availability. Senior Vice President Dave Treadwell emphasized the need for better AI usage practices and mandated senior engineer oversight.
