In recent news, Amazon Web Services (AWS) experienced a significant outage lasting 13 hours. This disruption was reportedly linked to one of its artificial intelligence tools, as covered by the Financial Times. The incident occurred in December when engineers implemented changes using the Kiro AI coding tool, according to sources familiar with the situation.
The Kiro tool functions autonomously, enabling it to perform actions on behalf of users. In this instance, it apparently decided to “delete and recreate the environment,” which led to the prolonged service interruption, primarily affecting operations in China.
Amazon claims that the involvement of AI tools was merely coincidental and maintains that the same situation could arise from any manual action or developer tool. According to the company, the outage was the result of “user error, not AI error.” They further noted that the Kiro tool typically requires authorization before proceeding with actions, but the employee involved had “broader permissions than expected,” indicating a user access control issue rather than an issue related to the tool’s autonomy.
Several employees at Amazon spoke with the Financial Times, revealing this incident was “at least” the second time in recent months that AI tools were implicated in service disruptions. “These outages were small but entirely predictable,” remarked one senior AWS employee.
Kiro was introduced by Amazon in July, with the company actively promoting its use among employees. Leadership has set a target for 80 percent weekly usage, closely monitoring how well the tool is adopted. Interestingly, Amazon also offers subscriptions to the Kiro tool.
These recent outages follow a more severe incident in October, which resulted in a 15-hour outage that affected platforms like Alexa, Snapchat, Fortnite, and Venmo. Amazon attributed that incident to a bug within its automation software.
Despite the disruptions, Amazon contests that certain products and services being unavailable should not necessarily be labeled as an outage. In response to the Financial Times report, the company provided the following statement:
We want to clarify the inaccuracies in the Financial Times’ reporting yesterday. The brief service interruption reported was due to user error—specifically, misconfigured access controls—and not AI as suggested in the article.
The disruption was an extremely limited event in December affecting a single service (AWS Cost Explorer, which assists customers in visualizing, understanding, and managing AWS expenses) in one of our 39 Geographic Regions globally. It did not impact compute, storage, database, AI technologies, or any other services we provide. The issue arose from a misconfigured role, similar to what could occur with any developer tool or manual action. We did not receive any customer inquiries regarding this minor interruption. We’ve implemented additional safeguards to prevent a recurrence—not because of a significant impact (which it wasn’t), but because we aim to learn from all our operational experiences to enhance security and resilience. Enhanced measures comprise mandatory peer reviews for production access. While incidents involving misconfigured access controls can happen with any developer tool, AI-powered or not, we believe it is vital to extract lessons from these experiences. The assertion that a second event affected AWS is entirely false.
For more than 20 years, Amazon has demonstrated high operational excellence through our Correction of Error (COE) process. We thoroughly review incidents so that we can learn and address issues before they escalate into larger problems.
Update, February 21, 2026, 11:58 AM ET: This article has been updated to include Amazon’s complete statement regarding the Financial Times report.
