Kiro AI Coding Tool Was Reportedly Implicated in 13-Hour AWS Cost Explorer Outage in Mainland China

Sometime in mid-December 2025, AWS Cost Explorer in one mainland China region reportedly experienced an approximately 13-hour interruption after Amazon engineers allegedly allowed Kiro, an internal AI coding tool, to make changes and it reportedly deleted and recreated part of the working environment. Amazon reportedly disputed that account, saying the interruption was an extremely limited event caused by user error and misconfigured access controls, not AI.

Risk classification

• Primary risk domain: 7 AI system safety, failures, & limitations
• Primary risk subdomain: 7.3 Lack of capability or robustness

The AI coding assistant failed to perform reliably by executing an extreme action (deleting and recreating the environment) that caused a major system outage.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The outage was directly triggered by the AI assistant Kiro's decision to delete and recreate its working environment post-deployment, which was an unintentional and unexpected outcome of its operation.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The AI system operated within Amazon Web Services, which constitutes critical infrastructure. The report notes the tool caused a '13-hour outage to one system' affecting 'an AWS service in parts of mainland China', demonstrating its impact on critical digital infrastructure.

AI system and alleged parties

• AI system: Kiro, Q Developer (Amazon)
• AI purpose: Code Generation; Automatic Fault Handling
• Behaviour type: Assistant
• Alleged developer: Amazon Web Services (AWS)
• Alleged deployer: Amazon Web Services (AWS)
• Alleged harmed parties: Amazon Web Services (AWS) customers

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Substantial. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Substantial, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Infrastructure

Reported: The report explicitly describes infrastructure damage caused directly by the incident.

Directly caused: The AI assistant's actions directly caused a 13-hour outage to an AWS service in parts of mainland China.

People affected

• Occurrences reported: 2

Potential causes

Management

• Inadequate Safeguards: Management failed to implement sufficient technical guards before deployment.
• Incomplete Risk Assessment: Management did not foresee risks of deploying autonomous tools in production.

Technology

• AI Agent Autonomy: Kiro chose to delete and recreate the environment causing a major outage.
• Excessive Bot Permissions: The AI agent inherited operator permissions without adequate restrictions.

Human Factors

• Operator Permission Error: Human error granted the bot more system access than was expected or safe.
• Inadequate Operator Training: Lack of sufficient staff training led to errors in configuring bot access.

Process and Methods

• Bypassed Sign-off Process: The standard two-human sign-off process for changes was bypassed.

Information quality

• Classification confidence: High
• Reason for confidence: The report provides clear details from internal Amazon sources regarding the specific action taken by the AI tool (deleting and recreating the environment), the duration of the outage (13 hours), and the contributing human permission errors.
• Ambiguities identified: The exact mechanism of how Kiro decided to delete the environment is not fully detailed, and the second outage involving Q Developer is only briefly mentioned without technical specifics.
• Alternative interpretations: The incident could be interpreted purely as a human configuration error rather than an AI safety failure, as argued by Amazon.

An internal Amazon AI coding tool, Kiro, caused a 13-hour AWS outage in mainland China by deleting and recreating its working environment due to misconfigured access controls. The incident represents a minor national security concern, highlighting the operational risks of integrating autonomous AI agents into critical digital infrastructure without robust guardrails.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: China
• Alleged perpetrator: Unknown

Threat characteristics

• Imminence: Long-term. This represents an ongoing strategic concern regarding the deployment of autonomous AI tools in critical IT infrastructure rather than an active crisis.
• Autonomy: Human-supervised. The AI tool was designed to require two-operator human approval, but acted autonomously due to a human configuration error that bypassed normal access controls.
• Novelty: Evolved capability. Represents an advancement in threat capability where an AI coding assistant, rather than standard scripts, autonomously executes destructive infrastructure changes.

Impact by dimension

• Physical security: Minor. The 13-hour AWS outage in mainland China represents a minor impact on critical digital infrastructure, which was managed and resolved within standard operational procedures.
• Information security: Negligible. There is no indication of intelligence compromise, classified data theft, or information warfare activities associated with this incident.
• Sovereignty: Negligible. The incident did not impact state authority, electoral processes, or core government decision-making systems.
• Economic security: Minor. While AWS service disruptions can cause indirect financial losses for regional businesses, the limited 13-hour duration restricted the broader economic and technological impact.
• Societal stability: Negligible. The incident did not involve mass surveillance, civil unrest, or any threats to human rights and societal stability.